{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:MOHU7SL5HLM5BGNAVXLPRI6CEA","short_pith_number":"pith:MOHU7SL5","schema_version":"1.0","canonical_sha256":"638f4fc97d3ad9d099a0add6f8a3c22022765c41dd3eecdd8580b5678323159c","source":{"kind":"arxiv","id":"2305.14555","version":1},"attestation_state":"computed","paper":{"title":"All Roads Lead to Rome? Exploring the Invariance of Transformers' Representations","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bernhard Sch\\\"olkopf, Mrinmaya Sachan, Qipeng Guo, Ryan Cotterell, Shauli Ravfogel, Yuxin Ren, Zhijing Jin","submitted_at":"2023-05-23T22:30:43Z","abstract_excerpt":"Transformer models bring propelling advances in various NLP tasks, thus inducing lots of interpretability research on the learned representations of the models. However, we raise a fundamental question regarding the reliability of the representations. Specifically, we investigate whether transformers learn essentially isomorphic representation spaces, or those that are sensitive to the random seeds in their pretraining process. In this work, we formulate the Bijection Hypothesis, which suggests the use of bijective methods to align different models' representation spaces. We propose a model ba"},"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":"2305.14555","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-23T22:30:43Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"b3f178f22b14e1c66f62c6dbbf3bed7ffe2d8873cf7a002e21e44922a7220dfc","abstract_canon_sha256":"aca1c2875b931a8403a5236cb72af838f221e86eb1a4713ad71b7569b6a6942b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:13:24.942974Z","signature_b64":"eGbVuCH6dXB8nxv7UN15k9K4aATj/ejwVkRj5z+JHkq/hL5Tgptu1tRG/sz54KVmin5il39a/rhDLFK+elM6Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"638f4fc97d3ad9d099a0add6f8a3c22022765c41dd3eecdd8580b5678323159c","last_reissued_at":"2026-07-05T06:13:24.942602Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:13:24.942602Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"All Roads Lead to Rome? Exploring the Invariance of Transformers' Representations","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bernhard Sch\\\"olkopf, Mrinmaya Sachan, Qipeng Guo, Ryan Cotterell, Shauli Ravfogel, Yuxin Ren, Zhijing Jin","submitted_at":"2023-05-23T22:30:43Z","abstract_excerpt":"Transformer models bring propelling advances in various NLP tasks, thus inducing lots of interpretability research on the learned representations of the models. However, we raise a fundamental question regarding the reliability of the representations. Specifically, we investigate whether transformers learn essentially isomorphic representation spaces, or those that are sensitive to the random seeds in their pretraining process. In this work, we formulate the Bijection Hypothesis, which suggests the use of bijective methods to align different models' representation spaces. We propose a model ba"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.14555","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/2305.14555/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":"2305.14555","created_at":"2026-07-05T06:13:24.942659+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.14555v1","created_at":"2026-07-05T06:13:24.942659+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.14555","created_at":"2026-07-05T06:13:24.942659+00:00"},{"alias_kind":"pith_short_12","alias_value":"MOHU7SL5HLM5","created_at":"2026-07-05T06:13:24.942659+00:00"},{"alias_kind":"pith_short_16","alias_value":"MOHU7SL5HLM5BGNA","created_at":"2026-07-05T06:13:24.942659+00:00"},{"alias_kind":"pith_short_8","alias_value":"MOHU7SL5","created_at":"2026-07-05T06:13:24.942659+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22202","citing_title":"Structure Retention in Embedding Spaces as a Predictor of Benchmark Performance","ref_index":119,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA","json":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA.json","graph_json":"https://pith.science/api/pith-number/MOHU7SL5HLM5BGNAVXLPRI6CEA/graph.json","events_json":"https://pith.science/api/pith-number/MOHU7SL5HLM5BGNAVXLPRI6CEA/events.json","paper":"https://pith.science/paper/MOHU7SL5"},"agent_actions":{"view_html":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA","download_json":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA.json","view_paper":"https://pith.science/paper/MOHU7SL5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.14555&json=true","fetch_graph":"https://pith.science/api/pith-number/MOHU7SL5HLM5BGNAVXLPRI6CEA/graph.json","fetch_events":"https://pith.science/api/pith-number/MOHU7SL5HLM5BGNAVXLPRI6CEA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA/action/storage_attestation","attest_author":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA/action/author_attestation","sign_citation":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA/action/citation_signature","submit_replication":"https://pith.science/pith/MOHU7SL5HLM5BGNAVXLPRI6CEA/action/replication_record"}},"created_at":"2026-07-05T06:13:24.942659+00:00","updated_at":"2026-07-05T06:13:24.942659+00:00"}