{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JVXMEBYE5TLU6NGBF6VU35VEQK","short_pith_number":"pith:JVXMEBYE","schema_version":"1.0","canonical_sha256":"4d6ec20704ecd74f34c12fab4df6a482ad196a0b3bd259da2bb77a7d41cc3a3d","source":{"kind":"arxiv","id":"2605.23591","version":1},"attestation_state":"computed","paper":{"title":"Asymmetric Scaling Laws from Sparse Features","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.LG","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"John Sous, Michael Winer","submitted_at":"2026-05-22T13:00:52Z","abstract_excerpt":"We introduce a model for neural scaling laws under sparse activations. In the model, test loss is often dominated by rare coordinates that are never observed in the training input. This mechanism induces a novel bottleneck absent from dense models. We derive the asymptotic population loss in both the underparameterized and overparameterized regimes, and show that the loss exhibits a double-descent peak near the interpolation threshold -- where the number of parameters is just sufficient to fit the training data -- resulting in a loss curve governed by two distinct scaling exponents -- one for "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.23591","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-22T13:00:52Z","cross_cats_sorted":["cond-mat.dis-nn","cs.LG","math.ST","stat.TH"],"title_canon_sha256":"50f7d3a0fe1bd72715a2bade92ea310cc9df4459a1b461d5ec6e6c2a3d8099e2","abstract_canon_sha256":"3c6720e54be800ea8e4bc76be514c8b48f5094ea8338ef9267ea831872fe93d3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:02:21.231691Z","signature_b64":"hHFAFsjFgEhF0Wa3DFuTO0mStfqGAQX+4eS3+D2RGOZRKMQIVLOucZCc17dcZZFmiEJRVOQlJnZ9UyX7HMDCBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4d6ec20704ecd74f34c12fab4df6a482ad196a0b3bd259da2bb77a7d41cc3a3d","last_reissued_at":"2026-05-25T02:02:21.230873Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:02:21.230873Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Asymmetric Scaling Laws from Sparse Features","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.LG","math.ST","stat.TH"],"primary_cat":"stat.ML","authors_text":"John Sous, Michael Winer","submitted_at":"2026-05-22T13:00:52Z","abstract_excerpt":"We introduce a model for neural scaling laws under sparse activations. In the model, test loss is often dominated by rare coordinates that are never observed in the training input. This mechanism induces a novel bottleneck absent from dense models. We derive the asymptotic population loss in both the underparameterized and overparameterized regimes, and show that the loss exhibits a double-descent peak near the interpolation threshold -- where the number of parameters is just sufficient to fit the training data -- resulting in a loss curve governed by two distinct scaling exponents -- one for "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.23591","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.23591/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.23591","created_at":"2026-05-25T02:02:21.230997+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.23591v1","created_at":"2026-05-25T02:02:21.230997+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.23591","created_at":"2026-05-25T02:02:21.230997+00:00"},{"alias_kind":"pith_short_12","alias_value":"JVXMEBYE5TLU","created_at":"2026-05-25T02:02:21.230997+00:00"},{"alias_kind":"pith_short_16","alias_value":"JVXMEBYE5TLU6NGB","created_at":"2026-05-25T02:02:21.230997+00:00"},{"alias_kind":"pith_short_8","alias_value":"JVXMEBYE","created_at":"2026-05-25T02:02:21.230997+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK","json":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK.json","graph_json":"https://pith.science/api/pith-number/JVXMEBYE5TLU6NGBF6VU35VEQK/graph.json","events_json":"https://pith.science/api/pith-number/JVXMEBYE5TLU6NGBF6VU35VEQK/events.json","paper":"https://pith.science/paper/JVXMEBYE"},"agent_actions":{"view_html":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK","download_json":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK.json","view_paper":"https://pith.science/paper/JVXMEBYE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.23591&json=true","fetch_graph":"https://pith.science/api/pith-number/JVXMEBYE5TLU6NGBF6VU35VEQK/graph.json","fetch_events":"https://pith.science/api/pith-number/JVXMEBYE5TLU6NGBF6VU35VEQK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK/action/storage_attestation","attest_author":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK/action/author_attestation","sign_citation":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK/action/citation_signature","submit_replication":"https://pith.science/pith/JVXMEBYE5TLU6NGBF6VU35VEQK/action/replication_record"}},"created_at":"2026-05-25T02:02:21.230997+00:00","updated_at":"2026-05-25T02:02:21.230997+00:00"}