{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:AP5HL6I5YSRKLEIZS4DUHJ35SF","short_pith_number":"pith:AP5HL6I5","schema_version":"1.0","canonical_sha256":"03fa75f91dc4a2a59119970743a77d914859e6ac3dfcaa6ed138f456a088e65a","source":{"kind":"arxiv","id":"2502.00264","version":2},"attestation_state":"computed","paper":{"title":"Beyond the Permutation Symmetry of Transformers: The Role of Rotation for Model Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Binchi Zhang, Jundong Li, Zaiyi Zheng, Zhengzhang Chen","submitted_at":"2025-02-01T01:44:55Z","abstract_excerpt":"Symmetry in the parameter space of deep neural networks (DNNs) has proven beneficial for various deep learning applications. A well-known example is the permutation symmetry in Multi-Layer Perceptrons (MLPs), where permuting the rows of weight matrices in one layer and applying the inverse permutation to adjacent layers yields a functionally equivalent model. While permutation symmetry fully characterizes the equivalence set for MLPs, its discrete nature limits its utility for transformers. In this paper, we introduce rotation symmetry, a novel form of parameter space symmetry for transformers"},"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":"2502.00264","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-01T01:44:55Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"f7d6aacf42cd0e74cac8aaa3b1964eb672ff926ab60c19e7048f745117fc4408","abstract_canon_sha256":"a3e62cc16afec64f438ab99ea5f869e7ff221507a55c5ba9cfe6ab2f11802694"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:40.287493Z","signature_b64":"6jXL0tM0stbSxXqFdtruotd60KxJdKLMWO2Fj9V6z2JojJCPNGTrtfq83Ijmqi0Ti+5VIcrTbNSpMQr5pTO7Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"03fa75f91dc4a2a59119970743a77d914859e6ac3dfcaa6ed138f456a088e65a","last_reissued_at":"2026-07-05T11:11:40.286932Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:40.286932Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond the Permutation Symmetry of Transformers: The Role of Rotation for Model Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Binchi Zhang, Jundong Li, Zaiyi Zheng, Zhengzhang Chen","submitted_at":"2025-02-01T01:44:55Z","abstract_excerpt":"Symmetry in the parameter space of deep neural networks (DNNs) has proven beneficial for various deep learning applications. A well-known example is the permutation symmetry in Multi-Layer Perceptrons (MLPs), where permuting the rows of weight matrices in one layer and applying the inverse permutation to adjacent layers yields a functionally equivalent model. While permutation symmetry fully characterizes the equivalence set for MLPs, its discrete nature limits its utility for transformers. In this paper, we introduce rotation symmetry, a novel form of parameter space symmetry for transformers"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.00264","kind":"arxiv","version":2},"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/2502.00264/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":"2502.00264","created_at":"2026-07-05T11:11:40.286988+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.00264v2","created_at":"2026-07-05T11:11:40.286988+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.00264","created_at":"2026-07-05T11:11:40.286988+00:00"},{"alias_kind":"pith_short_12","alias_value":"AP5HL6I5YSRK","created_at":"2026-07-05T11:11:40.286988+00:00"},{"alias_kind":"pith_short_16","alias_value":"AP5HL6I5YSRKLEIZ","created_at":"2026-07-05T11:11:40.286988+00:00"},{"alias_kind":"pith_short_8","alias_value":"AP5HL6I5","created_at":"2026-07-05T11:11:40.286988+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23607","citing_title":"Scaling Linear Mode Connectivity and Merging to Billion Parameter Pretrained Transformers","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12488","citing_title":"Anomalies in Neural Network Field Theory","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF","json":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF.json","graph_json":"https://pith.science/api/pith-number/AP5HL6I5YSRKLEIZS4DUHJ35SF/graph.json","events_json":"https://pith.science/api/pith-number/AP5HL6I5YSRKLEIZS4DUHJ35SF/events.json","paper":"https://pith.science/paper/AP5HL6I5"},"agent_actions":{"view_html":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF","download_json":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF.json","view_paper":"https://pith.science/paper/AP5HL6I5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.00264&json=true","fetch_graph":"https://pith.science/api/pith-number/AP5HL6I5YSRKLEIZS4DUHJ35SF/graph.json","fetch_events":"https://pith.science/api/pith-number/AP5HL6I5YSRKLEIZS4DUHJ35SF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF/action/storage_attestation","attest_author":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF/action/author_attestation","sign_citation":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF/action/citation_signature","submit_replication":"https://pith.science/pith/AP5HL6I5YSRKLEIZS4DUHJ35SF/action/replication_record"}},"created_at":"2026-07-05T11:11:40.286988+00:00","updated_at":"2026-07-05T11:11:40.286988+00:00"}