{"as_of":"2026-08-08T08:52:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:922543b2ba9e9f74ba666063edf508c0742a7cb7b7c0ab2d5ad16814adec132e","coverage":[{"denominator":73,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":73,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-07T00:43:09.868839Z","state":"measured"},{"denominator":75,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":75,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-08T06:32:00.761636+00:00","state":"measured"},{"denominator":2,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":2,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-01T14:42:34.657120Z","state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"arxiv_reference","source_observed_at":"2026-07-04T10:39:45.033475Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"cited_work":{"arxiv_id":"2506.13234","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2506.13234","snapshot_observed_at":"2026-07-04T10:39:45.033475Z","title":"The butterfly effect: Neu- ral network training trajectories are highly sensitive to initial conditions.arXiv preprint arXiv:2506.13234, 2025","venue":null,"work_id":"dd0f128a-7551-4e0b-8816-5c9701551b36","year":2025},"citing_paper":{"arxiv_id":"2606.23607","last_updated":"2026-06-22T17:08:28Z","snapshot_observed_at":"2026-08-05T16:18:42.526731Z","submitted_at":"2026-06-22T17:08:28Z","title":"Scaling Linear Mode Connectivity and Merging to Billion Parameter Pretrained Transformers","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-06-26T08:39:59.835392Z"},"links":{"cited_paper":"/paper/2506.13234","citing_paper":"/paper/2606.23607"},"observation_digest":"sha256:615e1f5cadb45cfc985e0539f65daef07ac23e6e1435b70931a6313bd71f880e","observation_id":"5addd247-0f30-4837-865d-ac08e2016a70","resolution":{"observed_at":"2026-07-04T10:39:45.035541Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2506.13234","snapshot_observed_at":"2026-08-01T14:42:34.657120Z","title":null,"venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2607.21637","last_updated":"2026-07-21T03:30:58Z","snapshot_observed_at":"2026-08-07T08:30:51.139786Z","submitted_at":"2026-07-21T03:30:58Z","title":"Quasi-Monte Carlo Initialization for Meta-Reinforcement Learning","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-01T14:42:34.657120Z"},"links":{"cited_paper":"/paper/2506.13234","citing_paper":"/paper/2607.21637"},"observation_digest":"sha256:322e848be72a9be6f3883d28af6f9bcaf563530bff0a7b14c0ecb206443d1db9","observation_id":"fc7d40c4-5438-430d-b000-d6116a2985c9","resolution":{"observed_at":"2026-08-01T14:42:34.657120Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"links":{"evidence":"/evidence","html":"/paper/2506.13234/citation-record","integrity":"/paper/2506.13234/integrity","json":"/paper/2506.13234/citation-record.json","paper":"/paper/2506.13234"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:05.100047Z","title":"write newline","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":1,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.100047Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:037ea6ac445e124951cb61baaabbc748d2c115a71ec229da953bc027de361e80","observation_id":"eb4a1a6b-df93-45ea-9a23-44c197a959d6","resolution":{"observed_at":"2026-08-07T00:43:05.100047Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.438701Z","title":"Layer-wise linear mode connectivity","venue":null,"work_id":"a04a1d3e-053d-4f37-bdf0-06cc055dfe20","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":2,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.157972Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:f9f547c286b0c133748e6e3f5f165cd5a8729ea0658b94c84871d25289124ffe","observation_id":"9455d820-76eb-4d57-a063-c9bc37aa954e","resolution":{"observed_at":"2026-08-07T00:43:13.442863Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.425156Z","title":"Git re-basin: Merging models modulo permutation symmetries","venue":null,"work_id":"0001d0b5-cdb2-46ba-939c-62294a60003c","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":3,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.210943Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:3e3deac5faf988716f44804f16dc51695d7b4c17385dd0420bc1929352218028","observation_id":"0d47d3df-3b99-495c-8745-4434ed4d09ad","resolution":{"observed_at":"2026-08-07T00:43:13.430227Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.412211Z","title":"S., Bachmann, G., Noci, L., and Hofmann, T","venue":null,"work_id":"4221dae1-d04c-4083-a1f9-0e7882e72bac","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":4,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.280830Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:79047a5b58ddfbf9c82c08acb2c0ad3d7d0d757fe01fea19596ac09bb49b69e9","observation_id":"33fc3d64-5ff8-4ce7-beb1-416cb4bf9be0","resolution":{"observed_at":"2026-08-07T00:43:13.416386Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"1607.06450","last_updated":"2016-07-21T19:57:52Z","snapshot_observed_at":"2026-07-06T02:11:23.670680Z","submitted_at":"2016-07-21T19:57:52Z","title":"Layer Normalization","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1607.06450","snapshot_observed_at":"2026-08-07T00:43:05.358766Z","title":"L., Kiros, J","venue":null,"work_id":null,"year":2016},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":5,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.358766Z"},"links":{"cited_paper":"/paper/1607.06450","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:09f5652837ed2a70ce8122565f844582f840002254e1cd2571a6d4587f0cda97","observation_id":"f07bd3ba-185b-4ae8-a73b-9f5d1d0f2f88","resolution":{"observed_at":"2026-08-07T00:43:05.358766Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.398591Z","title":"V., Akram, Y., Zucchet, N., Aitchison, L., and Steger, A","venue":null,"work_id":"c49555ed-1ffe-4f47-8186-c27dc24c96bc","year":2022},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":6,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.437808Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:98d1d364821751d29633479eeb7fc36931b96babe087187d10d24f17dc28b698","observation_id":"2ea6b26c-e97e-4067-b849-f42b9649ae33","resolution":{"observed_at":"2026-08-07T00:43:13.403319Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2108.09507","last_updated":"2022-07-27T21:56:42Z","snapshot_observed_at":"2026-08-05T21:02:33.708248Z","submitted_at":"2021-08-21T13:18:49Z","title":"Shift-Curvature, SGD, and Generalization","version":3},"cited_work":{"arxiv_id":"2108.09507","doi":null,"metadata_source":"pith","pith_arxiv_id":"2108.09507","snapshot_observed_at":"2026-08-07T00:43:10.738073Z","title":"Shift-Curvature, SGD, and Generalization","venue":"stat.ML","work_id":"6149dbbc-dbfa-4528-8922-249c829c194d","year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":7,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.505991Z"},"links":{"cited_paper":"/paper/2108.09507","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:995737065c072faff6deca0e5266b1405f24b7bfd0efc1fab5d4709640198b19","observation_id":"f2dd72fa-b764-4325-81f2-c213f5ffdd3b","resolution":{"observed_at":"2026-08-07T00:43:10.802107Z","resolver_source":"local_arxiv","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2110.14168","last_updated":"2021-11-18T00:23:45Z","snapshot_observed_at":"2026-08-07T01:45:38.840969Z","submitted_at":"2021-10-27T04:49:45Z","title":"Training Verifiers to Solve Math Word Problems","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2110.14168","snapshot_observed_at":"2026-08-07T00:43:05.652763Z","title":"Training verifiers to solve math word problems","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":9,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.652763Z"},"links":{"cited_paper":"/paper/2110.14168","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:8a8f86d4ea5d89b0177416994ee4de041214e09012f20b5fcd024c4f5f417185","observation_id":"d57522d1-d45e-4457-93bc-0d847f0af79f","resolution":{"observed_at":"2026-08-07T00:43:05.652763Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2103.00065","last_updated":"2022-11-23T18:09:57Z","snapshot_observed_at":"2026-08-06T02:48:08.712124Z","submitted_at":"2021-02-26T22:08:19Z","title":"Gradient Descent on Neural Networks Typically Occurs at the Edge of Stability","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2103.00065","snapshot_observed_at":"2026-08-07T00:43:05.706153Z","title":"M., Kaur, S., Li, Y., Kolter, J","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":10,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.706153Z"},"links":{"cited_paper":"/paper/2103.00065","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:c9d71605a1e7f5d15c9de4cec3e7f1995211a006cd0ec4001f9b54a84295aea3","observation_id":"6deacaab-fc79-46a1-8190-8030d5fbd5e1","resolution":{"observed_at":"2026-08-07T00:43:05.706153Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2310.04415","last_updated":"2024-11-04T23:21:13Z","snapshot_observed_at":"2026-07-06T16:28:55.209047Z","submitted_at":"2023-10-06T17:58:21Z","title":"Why Do We Need Weight Decay in Modern Deep Learning?","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2310.04415","snapshot_observed_at":"2026-08-07T00:43:05.785980Z","title":"Why do we need weight decay in modern deep learning?, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":11,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.785980Z"},"links":{"cited_paper":"/paper/2310.04415","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:415009f9d4a6dd4a07dd1b487205db49aaf375634e7232fe193b3a49e8632a3a","observation_id":"cae1a793-4f5b-45d6-94d1-29bbed05d785","resolution":{"observed_at":"2026-08-07T00:43:05.785980Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:05.840133Z","title":"BERT : Pre-training of deep bidirectional transformers for language understanding","venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":12,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.840133Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:00d27913f8e2738bf9e46dd57183ec8433a98b5b42fa5f0a9d453b0fb4f34860","observation_id":"77d0ed8a-39b4-47e0-8a28-996c0917c820","resolution":{"observed_at":"2026-08-07T00:43:05.840133Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:05.927440Z","title":"An image is worth 16x16 words: Transformers for image recognition at scale","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":13,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.927440Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:869406a8771d9694c0684ac138e9dd4e65525708de6e510e417d44fe7a1464c9","observation_id":"97ec0c4d-026f-4e71-959b-c8c5f9192cbd","resolution":{"observed_at":"2026-08-07T00:43:05.927440Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.376733Z","title":"Essentially No Barriers in Neural Network Energy Landscape","venue":null,"work_id":"53e723f2-7e26-4769-9f25-915fce9fe826","year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":14,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:05.974558Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:796931b39534bfdb9fdc439e9bd357574e4580bafbc7b642e44e4b2a6641fe8a","observation_id":"5576d2a3-c04d-4dbd-9544-f334136af34a","resolution":{"observed_at":"2026-08-07T00:43:13.380962Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.364372Z","title":"The role of permutation invariance in linear mode connectivity of neural networks","venue":null,"work_id":"c34f3472-1fb3-4961-b0f0-387237e54eb3","year":2022},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":15,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.021737Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:0c5a784f5e803dc32f4e4fe5c799f2030c3c0c9fc9b6b29b3520dbbd8f3ea239","observation_id":"ecd9b941-8d30-47ff-9d7b-3a26ea28d232","resolution":{"observed_at":"2026-08-07T00:43:13.368385Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.351048Z","title":"Deep ensembles: A loss landscape perspective, 2019","venue":null,"work_id":"ba7154c9-e172-4513-9acb-d0b7dd2daeef","year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":16,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.109966Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:81c18ef7020b30cbe92a63761d9e139eb5450d64ef2085ae6632cef3c43e05e8","observation_id":"bc154909-b8b3-4591-841b-f9bde4410cbb","resolution":{"observed_at":"2026-08-07T00:43:13.355836Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:06.167369Z","title":"K., Paul, M., Kharaghani, S., Roy, D","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":17,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.167369Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:39be220df0addf0cd75153c40bf49e07d7e488d96f94c24c1bef05583faba95a","observation_id":"c1c54ca5-d3e2-48ed-8b50-aaea4eec6b04","resolution":{"observed_at":"2026-08-07T00:43:06.167369Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.338001Z","title":"and Carbin, M","venue":null,"work_id":"cf8495df-9961-4f5a-b58d-92058525ea5f","year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":18,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.224950Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:809b829cf259ebfdf8c167fc5974acda98d27c16a6b62e1d8e24dc8711ecd467","observation_id":"a2b4c10b-c885-4911-ba48-7379fee3a0ba","resolution":{"observed_at":"2026-08-07T00:43:13.342084Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.326076Z","title":"K., Roy, D., and Carbin, M","venue":null,"work_id":"a48acc84-6cbb-4864-948f-ee57dcecefb2","year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":19,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.273531Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:b1f7645898780f572c869046236dc531d3d416dfeeb8d9560b1a15cc185106c2","observation_id":"ea1184bf-b8c5-4632-a068-56b4db380d37","resolution":{"observed_at":"2026-08-07T00:43:13.329733Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.310471Z","title":"J., and Morcos, A","venue":null,"work_id":"053f84c7-7b1b-454c-b2a3-90c151e04533","year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":20,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.337408Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:5be23a7affc2fa0073d681be989984f6ebf077e448dad57d42f7eca0f88d7706","observation_id":"2c9a9f62-50f6-4083-9476-8159135a0b5f","resolution":{"observed_at":"2026-08-07T00:43:13.317064Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.297974Z","title":"P., and Wilson, A","venue":null,"work_id":"a2bfa14e-bbed-49fa-9eeb-01b4ca131b1f","year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":21,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.414070Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:bfdb7abfb5f91038f02f59abdbcd864184be59d0d7281b0348e7a45469cdc63f","observation_id":"36d495ba-d25d-4641-831e-2804117bc2c3","resolution":{"observed_at":"2026-08-07T00:43:13.302005Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"1412.6544","last_updated":"2015-05-21T21:44:31Z","snapshot_observed_at":"2026-07-06T04:04:16.777653Z","submitted_at":"2014-12-19T21:55:01Z","title":"Qualitatively characterizing neural network optimization problems","version":6},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1412.6544","snapshot_observed_at":"2026-08-07T00:43:06.474225Z","title":"J., Vinyals, O., and Saxe, A","venue":null,"work_id":null,"year":2015},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":22,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.474225Z"},"links":{"cited_paper":"/paper/1412.6544","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:e32ee0f877bbb60be68a628b4752dbe8d9f86b5f8e213c38e36851be839eabd7","observation_id":"bd56b550-3ce3-463a-a077-a0e3fa02fde6","resolution":{"observed_at":"2026-08-07T00:43:06.474225Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2402.00838","last_updated":"2024-06-07T21:59:52Z","snapshot_observed_at":"2026-07-06T17:23:51.547578Z","submitted_at":"2024-02-01T18:28:55Z","title":"OLMo: Accelerating the Science of Language Models","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2402.00838","snapshot_observed_at":"2026-08-07T00:43:06.553941Z","title":"H., Ivison, H., Magnusson, I., Wang, Y., Arora, S., Atkinson, D., Authur, R., Chandu, K","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":23,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.553941Z"},"links":{"cited_paper":"/paper/2402.00838","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:dca1a4b38c1afc333f6afcbb94830dc2dfa30e5fe35f3e75e3aba4943a7856ef","observation_id":"20f28c73-28f5-4741-a562-6a58337b62a8","resolution":{"observed_at":"2026-08-07T00:43:06.553941Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.285551Z","title":"Delving deep into rectifiers: Surpassing human-level performance on imagenet classification","venue":null,"work_id":"3f2a1f1a-f9e1-42bb-a35b-977321cba125","year":2015},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":24,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.603814Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:ac374e9af699c7fbfbe00e9536b5be330994c8d406e256d0791d51e68d48008a","observation_id":"0e32780e-c0f6-4fa8-a367-b5b479cc4613","resolution":{"observed_at":"2026-08-07T00:43:13.289431Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:06.686925Z","title":"Deep residual learning for image recognition","venue":null,"work_id":null,"year":2016},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":25,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.686925Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:0119f73fbc817469fc21ca29e28f621768c18ef31ff0674e7264fae9ba37eac7","observation_id":"09cab4be-4db1-47c5-a25a-01b62d84faa9","resolution":{"observed_at":"2026-08-07T00:43:06.686925Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.263483Z","title":"E., and Weinberger, K","venue":null,"work_id":"6ff1e4bb-c7e3-4d0c-b1da-46521f4b24f8","year":2017},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":26,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.727623Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:8404a190da6b1bf5a1906503e430af4865c4b2ab7e4a768e46575c866fc61100","observation_id":"25a5439d-49ea-4a25-a5bb-4fcd7389b0e4","resolution":{"observed_at":"2026-08-07T00:43:13.267568Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.251078Z","title":"T., Wortsman, M., Schmidt, L., Hajishirzi, H., and Farhadi, A","venue":null,"work_id":"e52c24c5-84fe-4ca5-96ad-5107dd845809","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":27,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.780031Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:bf60970903b6e701015b7a290f9dab244ad531682225149534f1729edc362bae","observation_id":"03d99161-8650-45e4-833e-9b88a1bfdf8d","resolution":{"observed_at":"2026-08-07T00:43:13.255279Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.239141Z","title":"Analysis of linear mode connectivity via permutation-based weight matching","venue":null,"work_id":"db8269cf-7273-4f53-ae0b-8d722e8797df","year":2025},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":28,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.829610Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:4cd19839776379520b687c41771bd97e23ef6a77fb0597642652fb4e86be8dff","observation_id":"417228dd-0e58-4cf0-b78c-469f8fdcd366","resolution":{"observed_at":"2026-08-07T00:43:13.243071Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.227172Z","title":"Maximal initial learning rates in deep R e LU networks","venue":null,"work_id":"884c7458-3157-4b82-b149-d268b1639a05","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":29,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.886009Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:95c276f145d03bff17877f0e1c3a59e463f6f022194b678394cd968d08f9323c","observation_id":"cabbe003-5788-42ce-9442-0e06f1dc2a04","resolution":{"observed_at":"2026-08-07T00:43:13.231249Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.215269Z","title":"Averaging weights leads to wider optima and better generalization","venue":null,"work_id":"171c095c-6006-4c55-b103-7622c782f983","year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":30,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.935259Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:4c8f2d71ea226ee37f857f669c759c020bdd3e07493335013d8da883b8e11052","observation_id":"3a9f5bda-6533-471a-80e1-37daab5a6d86","resolution":{"observed_at":"2026-08-07T00:43:13.219123Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.203555Z","title":"Neural tangent kernel: Convergence and generalization in neural networks","venue":null,"work_id":"59095fb2-1c17-4cf2-a85f-005b7eaf0d81","year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":31,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:06.990533Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:8478423a5046a5e71e815b595c390d44c62e115104ce168c48d66eace573efdd","observation_id":"e7dfa132-d921-48d4-9040-ad169610af87","resolution":{"observed_at":"2026-08-07T00:43:13.207403Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.190416Z","title":"The break-even point on optimization trajectories of deep neural networks","venue":null,"work_id":"714d90e1-69f5-40ea-a355-13ea01fe5488","year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":32,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.044517Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:7a3126e780851cbfda3acbb1ad92fead1446869a54a58404142699458c9b6ca3","observation_id":"91ec5cfe-c0a0-4f4d-a628-93341303a18c","resolution":{"observed_at":"2026-08-07T00:43:13.194184Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.178211Z","title":"REPAIR : RE normalizing P ermuted A ctivations for I nterpolation R epair","venue":null,"work_id":"527496c5-c060-48a7-8b1f-4fe7f4df4936","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":33,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.106594Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:e56aaefecc82805ed0ae13d48fe5905ac7398267eaa8ecaa893812e4dbf67756","observation_id":"5680b034-4ad2-4ca5-a62c-8d22557b3b94","resolution":{"observed_at":"2026-08-07T00:43:13.182327Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.164371Z","title":"Linear connectivity reveals generalization strategies","venue":null,"work_id":"56dc4524-1fe7-447e-8856-f75961c7a7e7","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":34,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.137331Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:b4985184cd4aa0de0abb9ff0c96dd8b049be46d49ace09b0130030976ba7d69d","observation_id":"ffe8922c-d01a-4f71-8fc0-17e1916f52f6","resolution":{"observed_at":"2026-08-07T00:43:13.170161Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.151673Z","title":"S., Mudigere, D., Nocedal, J., Smelyanskiy, M., and Tang, P","venue":null,"work_id":"5e0ce234-132b-4297-9501-60e5b50b9816","year":2017},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":35,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.196965Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:f297b24abf7fc32c8ec83608d1feaa04d8481702b62c426590e0bc6d4359b8fa","observation_id":"a222b76a-489c-4bab-8efa-d1d941d2813c","resolution":{"observed_at":"2026-08-07T00:43:13.155578Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.139370Z","title":"Similarity of neural network representations revisited","venue":null,"work_id":"2f0e9880-ac63-4c1d-9b2e-a412571d2803","year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":36,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.249879Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:fcf2e8aae75c7fd0669fb0300fbba1eb5a08b4a959026f89b8f1c7439d9bdf26","observation_id":"d45fd1e2-f2c3-4175-8297-26281d2fcfa1","resolution":{"observed_at":"2026-08-07T00:43:13.143117Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.127142Z","title":"Learning multiple layers of features from tiny images, 2009","venue":null,"work_id":"c4a65a3b-3803-49ff-b5be-7a84900b5a54","year":2009},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":37,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.285654Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:bfe3cc5a06fbb58576c33a992c0304b4517315070e7d8dee935d70d0bafe0f9a","observation_id":"8eac8671-685a-42b2-9939-515ecee88f0d","resolution":{"observed_at":"2026-08-07T00:43:13.131122Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.115109Z","title":"D., Kwok, D., Matelsky, J","venue":null,"work_id":"cbe5922c-d5b0-4dfc-a07f-b2d9fdad8adc","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":38,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.341020Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:542a9c966216060cd80602196a1387b18270c4f0de118896db27f5ae0e923f67","observation_id":"1e8b9c71-3531-484c-8fe4-eb7c2113075b","resolution":{"observed_at":"2026-08-07T00:43:13.118603Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.102803Z","title":"Wide neural networks of any depth evolve as linear models under gradient descent","venue":null,"work_id":"15c7b731-aab6-4d95-bd8f-7854e4a9b6a2","year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":39,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.393213Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:4b9194af95f26957aadc4afad079bfdec20a36df8e4f5180f12a82f9c5523fad","observation_id":"77247a5f-663b-4c50-9df5-5cdbafd649b1","resolution":{"observed_at":"2026-08-07T00:43:13.106540Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2404.06391","last_updated":"2024-04-09T15:35:02Z","snapshot_observed_at":"2026-07-06T17:57:48.153936Z","submitted_at":"2024-04-09T15:35:02Z","title":"Exploring Neural Network Landscapes: Star-Shaped and Geodesic Connectivity","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2404.06391","snapshot_observed_at":"2026-08-07T00:43:07.446769Z","title":"Exploring neural network landscapes: Star-shaped and geodesic connectivity","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":40,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.446769Z"},"links":{"cited_paper":"/paper/2404.06391","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:c46210b6be994dbfb304330da52846142b268f0c37d558b38784d94716caf5a5","observation_id":"e4804cfb-3121-4c0b-8e13-09f1381aeb03","resolution":{"observed_at":"2026-08-07T00:43:07.446769Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.090343Z","title":"How good is a single basin? In Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, volume 238, pp.\\ 4015--4023","venue":null,"work_id":"65a14dc0-4c05-4f08-8b2d-c807ed89089f","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":41,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.499300Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:67bd793402992e4db480280c3af03bd66ea113d72d428ad74c8ab5fead68f358","observation_id":"52ca1ad0-e19e-4a3a-ac89-c5e8fbcaf0ae","resolution":{"observed_at":"2026-08-07T00:43:13.094633Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"1711.05101","last_updated":"2019-01-04T21:01:49Z","snapshot_observed_at":"2026-07-06T02:11:23.670680Z","submitted_at":"2017-11-14T14:24:06Z","title":"Decoupled Weight Decay Regularization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1711.05101","snapshot_observed_at":"2026-08-07T00:43:07.536563Z","title":"and Hutter, F","venue":null,"work_id":null,"year":2017},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":42,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.536563Z"},"links":{"cited_paper":"/paper/1711.05101","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:915c84e241019631d23321b21738aeade988151608b166f0b7b38d4b69bd2105","observation_id":"ccce7def-b85f-42b9-9545-078673ddb725","resolution":{"observed_at":"2026-08-07T00:43:07.536563Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.077607Z","title":"S., Bigelow, E","venue":null,"work_id":"b419e4a8-c64b-44f5-a9e8-1f3db509b33e","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":43,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.590518Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:f016659a66bf913b352209543928c67afa226e8516b821c3c3741d68a28b6642","observation_id":"e09601d7-8bf7-4ed0-a19e-22621c0ef130","resolution":{"observed_at":"2026-08-07T00:43:13.081529Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.063175Z","title":"I., Farajtabar, M., Gorur, D., Pascanu, R., and Ghasemzadeh, H","venue":null,"work_id":"bd976591-63c2-4bc8-88e0-833f126de915","year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":44,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.643827Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:47c38a5ca4930a53b0e5ab0e87c7a9e0d0ac6c8650e3220077378b03c42c100d","observation_id":"d530d861-cfa0-4c6b-8b64-ff174e2c618f","resolution":{"observed_at":"2026-08-07T00:43:13.067325Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.045556Z","title":"Equivariant deep weight space alignment","venue":null,"work_id":"5bcb49e0-a561-4232-a53c-d644f0352a48","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":45,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.673529Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:4d884a14eb3247320bc293f2ebf8578d55bb82d32a02c2afd1eae0cf582c590c","observation_id":"5743837a-51b9-4183-b8fc-56d03a1edcd6","resolution":{"observed_at":"2026-08-07T00:43:13.049517Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.032357Z","title":"What is being transferred in transfer learning? In Advances in Neural Information Processing Systems, volume 33, pp.\\ 512--523, 2020","venue":null,"work_id":"a4fcef6f-093b-4509-9a52-f9b38132aa23","year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":46,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.722230Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:8119fccfd23ad65108a86aad46c0238ce5262c21c7b53e95f2634dce6075852f","observation_id":"74629502-a691-494e-ab26-0172285ff859","resolution":{"observed_at":"2026-08-07T00:43:13.036674Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.018636Z","title":"Do wide and deep networks learn the same things? Uncovering how neural network representations vary with width and depth","venue":null,"work_id":"9290e847-5b62-4460-8e0e-9a9d7a46bed6","year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":47,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.780227Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:5c2bc19c2c2ae86d5831f7542baab6060da4535356b5fa3b07c97110ed21b8c5","observation_id":"dc310133-757f-4c46-9bf9-c20e61825e91","resolution":{"observed_at":"2026-08-07T00:43:13.023113Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:13.005051Z","title":null,"venue":null,"work_id":"a980efa1-ca4f-4254-bf29-6e5abba8dfc2","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":48,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.830971Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:281c287068de41ac7f52800798df7700163f7c67e0098445b27c1978ff471e44","observation_id":"01cbc6ee-6236-4bb2-82b0-5fe47f372231","resolution":{"observed_at":"2026-08-07T00:43:13.009630Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.991865Z","title":"SVCCA : Singular vector canonical correlation analysis for deep learning dynamics and interpretability","venue":null,"work_id":"67e2fa6d-e4ee-4bb0-8127-b32cc9abceb0","year":2017},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":49,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.882612Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:dbdf72dc27c3ed88873b154db3b6f3a24fe3ad7b2d96291c0aa5ce850a35a553","observation_id":"b2a2efc0-63fc-4835-b443-bed057b68b29","resolution":{"observed_at":"2026-08-07T00:43:12.995802Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.980646Z","title":"T., Bello-Rivas, J","venue":null,"work_id":"b75cfd89-2335-4a0a-b9d4-32405edd5af6","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":50,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.931699Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:a282bcaacd68a963ff74fb583eb8e6accc6c3c57a286229e25545c6e4d3eb528","observation_id":"1cd3e859-ba27-4f3c-a226-9d4acaccb2e0","resolution":{"observed_at":"2026-08-07T00:43:12.984018Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:07.968408Z","title":"C., and Fei-Fei, L","venue":null,"work_id":null,"year":2015},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":51,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:07.968408Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:e1efecd0aac9250b64a9ae36becd6b2e7b843b28cd773719ac9d6977e01661e0","observation_id":"c7eb8cbe-1ba9-467a-8418-e42f50c2d1a2","resolution":{"observed_at":"2026-08-07T00:43:07.968408Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.968602Z","title":"P., and Lobacheva, E","venue":null,"work_id":"4fa4a98a-0a0b-4330-8785-0eb00275426d","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":52,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.056120Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:406a5a447c44e032ef1cc6ea6637aa17c67d919925029591b6b0b54832900d56","observation_id":"42b97185-5501-4073-a8d4-7fa89fa81198","resolution":{"observed_at":"2026-08-07T00:43:12.972028Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.955731Z","title":"Random teachers are good teachers","venue":null,"work_id":"15c4503d-c297-4f17-9a7d-7240f9817876","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":53,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.123496Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:f758c8e6344753b5a2ede9a565783432a05871fee89f0ea42fb75163f2ad0d9f","observation_id":"1053087f-ff0e-4a9c-be33-be7f95d33d93","resolution":{"observed_at":"2026-08-07T00:43:12.960005Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.941448Z","title":"The MultiBERTs : BERT reproductions for robustness analysis","venue":null,"work_id":"da59ab4c-1ac7-46ef-95d9-ea103e682703","year":2022},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":54,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.173484Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:b6ea70ccb824e52dd353348adefbce4d361db4c7216f0833d52492052af055ff","observation_id":"2e255c62-1ab0-4453-a5c7-db42754a467a","resolution":{"observed_at":"2026-08-07T00:43:12.946040Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.928512Z","title":"M., Rolnick, D., and Dziugaite, G","venue":null,"work_id":"8fcb1ed9-d735-4f33-ae74-8d800a97a158","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":55,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.254228Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:8d0d2d93394437369718086672c657a8044f163b05b85456e601f76993b9d334","observation_id":"5ed3dca2-8616-4a1b-b413-4c8f48e93571","resolution":{"observed_at":"2026-08-07T00:43:12.932301Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.916525Z","title":"Geometry of the loss landscape in overparameterized neural networks: Symmetries and invariances","venue":null,"work_id":"a6be2686-5b10-408e-8fae-9f6f3284b215","year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":56,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.329096Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:c20f24291efca796f63c834c409439c78e07c72ea33d03ae26cf837305af8c5b","observation_id":"a588ae60-8204-425c-95ee-ea988037edef","resolution":{"observed_at":"2026-08-07T00:43:12.920243Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.758703Z","title":null,"venue":null,"work_id":"9fa30439-90e0-43f4-a54b-40ce38c33dfe","year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":57,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.406389Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:4ff78e5e941654c233baf0a8aa71bf3fa4fc20e5b913649ec6575238580c4374","observation_id":"c654e2c6-6839-425e-a11b-fe6189ad625d","resolution":{"observed_at":"2026-08-07T00:43:12.887670Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.627878Z","title":"P., Adilova, L., Kamp, M., Fischer, A., Schölkopf, B., Tübingen, M.-I., Hofmann, T., and Ch, E","venue":null,"work_id":"f82cef37-8b6f-4ed0-8ee0-e58624aab2e5","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":58,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.452867Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:6f1487157355785ddfa8757bc0dde0f180f2b6af1c6c30de8c83b5f98da734ee","observation_id":"984c3f57-b918-4bca-ad22-68600c4ececd","resolution":{"observed_at":"2026-08-07T00:43:12.722674Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"1803.09820","last_updated":"2018-04-24T17:43:51Z","snapshot_observed_at":"2026-07-06T06:30:19.251357Z","submitted_at":"2018-03-26T20:05:59Z","title":"A disciplined approach to neural network hyper-parameters: Part 1 -- learning rate, batch size, momentum, and weight decay","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1803.09820","snapshot_observed_at":"2026-08-07T00:43:08.528757Z","title":null,"venue":null,"work_id":null,"year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":59,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.528757Z"},"links":{"cited_paper":"/paper/1803.09820","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:ece462fa6889c9e08a88858d765b7b86adf8837db2b9829572169d426bedc469","observation_id":"b198fd57-bda2-4a97-9fb8-de7ac37083fe","resolution":{"observed_at":"2026-08-07T00:43:08.528757Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.499277Z","title":null,"venue":null,"work_id":"f0363a45-5cc2-4449-9f49-2ea645c00e1d","year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":60,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.591518Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:a08e5e2486cc782c736bc80a3528d7cd77c9d0709f9e1e9644260a8d71f23340","observation_id":"2989b5ac-13d6-4cfa-961f-be28b8d640d5","resolution":{"observed_at":"2026-08-07T00:43:12.557066Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2402.06184","last_updated":"2024-02-09T04:46:48Z","snapshot_observed_at":"2026-08-04T16:19:30.614896Z","submitted_at":"2024-02-09T04:46:48Z","title":"The boundary of neural network trainability is fractal","version":1},"cited_work":{"arxiv_id":"2402.06184","doi":null,"metadata_source":"pith","pith_arxiv_id":"2402.06184","snapshot_observed_at":"2026-08-07T00:43:10.358066Z","title":"The boundary of neural network trainability is fractal","venue":"cs.LG","work_id":"86ab72df-55de-401b-abf0-2915a05f88a0","year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":61,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.680044Z"},"links":{"cited_paper":"/paper/2402.06184","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:214050369fc774847136512f708174337d8a1a1800c6b96afb539e6e42b74b49","observation_id":"ac025386-8e78-47e3-b420-663ccce4141c","resolution":{"observed_at":"2026-08-07T00:43:10.423646Z","resolver_source":"local_arxiv","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2403.07968","last_updated":"2024-06-09T11:51:03Z","snapshot_observed_at":"2026-07-06T17:43:31.874919Z","submitted_at":"2024-03-12T13:59:23Z","title":"Do Deep Neural Network Solutions Form a Star Domain?","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2403.07968","snapshot_observed_at":"2026-08-07T00:43:08.736723Z","title":null,"venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":62,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.736723Z"},"links":{"cited_paper":"/paper/2403.07968","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:fd50b1c2fe17a6ff60947152b614e00ab89eed229b31427182d559fcd6d932f4","observation_id":"a05d770b-7108-4b4c-9d38-e3205fdff847","resolution":{"observed_at":"2026-08-07T00:43:08.736723Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2503.19206","last_updated":"2025-03-28T02:10:05Z","snapshot_observed_at":"2026-08-07T16:39:47.680195Z","submitted_at":"2025-03-24T23:11:56Z","title":"Overtrained Language Models Are Harder to Fine-Tune","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2503.19206","snapshot_observed_at":"2026-08-07T00:43:08.795214Z","title":"M., Goyal, S., Wen, K., Kumar, T., Yue, X., Malladi, S., Neubig, G., and Raghunathan, A","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":63,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.795214Z"},"links":{"cited_paper":"/paper/2503.19206","citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:4bc28ec5a1dbdc6206c84818f14e393ca733767c0d32b5f011e4decfca5b636b","observation_id":"0e75bfbd-d18c-48e7-901c-3750d973c91e","resolution":{"observed_at":"2026-08-07T00:43:08.795214Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.356980Z","title":"Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and engineering","venue":null,"work_id":"34628ece-d43d-45b7-a620-ebfc2f6047ed","year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":64,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.838912Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:1f0e229b31736df8f8e99107cfe0bae3b290dd8613a90f85970acdbdb95d730d","observation_id":"73027bf7-1fbd-4c1b-83dd-a81844519b0d","resolution":{"observed_at":"2026-08-07T00:43:12.412655Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.223347Z","title":"W., Thiery, A","venue":null,"work_id":"b0bb101f-61e6-4aee-be08-01f05d6460bc","year":2016},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":65,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:08.961688Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:b5496cab4f36fa225dfc56dabcae6ef095c23975bed265f51122f60363eebe6c","observation_id":"e110fd5b-ff1b-4c39-aa9d-0ccabe5f9352","resolution":{"observed_at":"2026-08-07T00:43:12.261679Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:12.072104Z","title":"Weight averaging for neural networks and local resampling schemes","venue":null,"work_id":"c9247bda-a99e-49d3-b668-40d6810e0b0f","year":1996},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":66,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.050894Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:143c2b14ba34ceafac2fca74c59eb518db68d9a9a3fc161d25e1ad5bfd582ab1","observation_id":"f13f49b5-fcd5-4786-a3b2-f6a38ac3d910","resolution":{"observed_at":"2026-08-07T00:43:12.171905Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:11.887627Z","title":null,"venue":null,"work_id":"e74a76fc-ec50-49a9-a029-d2f67feb22a1","year":2022},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":67,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.108545Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:98cae9dc795971cac9a7b5c23f06f04a4863d610267231ede7ff4f4b6e7457da","observation_id":"8f2e119f-7991-470a-8063-67d83a6802af","resolution":{"observed_at":"2026-08-07T00:43:11.934338Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:11.732367Z","title":null,"venue":null,"work_id":"2452c758-0801-46f2-abcd-fffeed543cbb","year":2019},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":68,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.223821Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:8883b8324f191cf37bb8263269a74ed46839795c0538533adb2dc34574023601","observation_id":"088c81ff-4834-4b17-a765-a5456a330f94","resolution":{"observed_at":"2026-08-07T00:43:11.799216Z","resolver_source":"raw_fallback","status":"unresolved"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:11.491910Z","title":"Federated learning with matched averaging","venue":null,"work_id":"f498a225-1bec-47f5-b743-fc595f9b2209","year":2020},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":69,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.357279Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:39d4499a05355f7d3650a87aa11481742348a59535e5f8800345b575043773e6","observation_id":"76977578-87c1-4165-84a8-62453b27abc3","resolution":{"observed_at":"2026-08-07T00:43:11.617288Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:11.326448Z","title":"H., Kunz, E., Kornblith, S., and Linderman, S","venue":null,"work_id":"cf6214be-039f-44cd-8dc0-ff4e8384b8fe","year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":70,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.453626Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:805ecfcd53df8fcf51ee21fc21e72dc48961da45d6b3c6ce3f792a21f3cd95d3","observation_id":"feeb1d3b-582d-4614-aecf-b5d49675d1da","resolution":{"observed_at":"2026-08-07T00:43:11.383302Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:11.173844Z","title":"C., Guestrin, C., Farhadi, A., and Rastegari, M","venue":null,"work_id":"2ee34d9a-9948-494d-b4da-7a1a91ec2087","year":2021},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":71,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.593002Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:52a5e95e0ccc37c3cc7d8467c13dfc29760915fa36e63e8790c59bc891a724b2","observation_id":"0ec77d66-531f-43b0-be75-bbf4cb18a167","resolution":{"observed_at":"2026-08-07T00:43:11.210564Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":"file/6651526","doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:10.138440Z","title":"How SGD selects the global minima in over-parameterized learning: A dynamical stability perspective","venue":null,"work_id":"88f0dc4a-ad47-466c-8e37-6c8980af762f","year":2018},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":72,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.706343Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:5c22012d8fba5ab62ef2eff4df25270bb74930d85d517063ec0e089bca0580ca","observation_id":"4ee66f01-88a7-4531-894f-5f08e7385fb4","resolution":{"observed_at":"2026-08-07T00:43:10.236794Z","resolver_source":"raw_fallback","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:11.025310Z","title":"K., Savarese, P","venue":null,"work_id":"abed4d66-99b2-43be-a2e4-3824da513959","year":2022},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":73,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.766735Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:b4aa9781c72cf18fe6f9bb9c73c006c6f8b664b07b38a9d1eaba87c1c9ef788c","observation_id":"cdefd4ab-d017-4ebd-927d-24a6b0888673","resolution":{"observed_at":"2026-08-07T00:43:11.068269Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T00:43:10.904613Z","title":"Going beyond linear mode connectivity: The layerwise linear feature connectivity","venue":null,"work_id":"321bd437-f40d-40c3-b092-e48487eadd89","year":2023},"citing_paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions","version":1},"reference_index":74,"source":"arxiv_source","source_observed_at":"2026-08-07T00:43:09.868839Z"},"links":{"citing_paper":"/paper/2506.13234"},"observation_digest":"sha256:c3e824b55a0940f36b2a8be02a7b7ee0ca7e48e469cbce1345349e3a978d3727","observation_id":"59fa170a-956e-4fdd-bfc4-b2f3452479b8","resolution":{"observed_at":"2026-08-07T00:43:10.984370Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-08T06:32:00.761636+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"state":"measured"}}],"paper":{"arxiv_id":"2506.13234","last_updated":"2025-06-16T08:35:16Z","latest_version":1,"primary_category":"cs.LG","snapshot_observed_at":"2026-08-07T00:34:22.881688Z","submitted_at":"2025-06-16T08:35:16Z","title":"The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions"},"reference_resolution":{"displayed":73,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":22,"verified_exact":3,"verified_fuzzy":48},"total_outbound_references":73},"refusal":"A citation records a reference. It does not transfer a finding from one paper to another.","schema":"pith.paper-citation-record.v1","standing_sources":[{"observed_at":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"thesis":"As of 8 August 2026, this Paper Citation Record lists 73 of 73 outbound references and 2 inbound Pith citation observations for arXiv:2506.13234."}