{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LRQGXTEMNQ7WSTVTGYYMFB4P23","short_pith_number":"pith:LRQGXTEM","schema_version":"1.0","canonical_sha256":"5c606bcc8c6c3f694eb33630c2878fd6c4eedd0061b8914ae0ca80775c13d8a7","source":{"kind":"arxiv","id":"2502.13963","version":1},"attestation_state":"computed","paper":{"title":"MuDAF: Long-Context Multi-Document Attention Focusing through Contrastive Learning on Attention Heads","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongmei Zhang, Ming Gong, Ning Wu, Shining Liang, Shiping Yang, Weihao Liu, Wenbiao Ding","submitted_at":"2025-02-19T18:59:15Z","abstract_excerpt":"Large Language Models (LLMs) frequently show distracted attention due to irrelevant information in the input, which severely impairs their long-context capabilities. Inspired by recent studies on the effectiveness of retrieval heads in long-context factutality, we aim at addressing this distraction issue through improving such retrieval heads directly. We propose Multi-Document Attention Focusing (MuDAF), a novel method that explicitly optimizes the attention distribution at the head level through contrastive learning. According to the experimental results, MuDAF can significantly improve the "},"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":"2502.13963","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-19T18:59:15Z","cross_cats_sorted":[],"title_canon_sha256":"ea2923204b8572033cd7ad3ea2dfb6ad3746ffcdda35fd267398c5370a1e4c82","abstract_canon_sha256":"5b6bee8fa2ee1b0d4f8e358dabb323e53cbbde9878c8f5067617c2354843b623"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:03.131969Z","signature_b64":"67pEzI1pCq4cvsWlFuDsacW5H/mwT3f1Q0C8vbFjmTyIdfttBiVM1cAN3RuM0yhsCfAJsyzZr72nhoQ8lP+EAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5c606bcc8c6c3f694eb33630c2878fd6c4eedd0061b8914ae0ca80775c13d8a7","last_reissued_at":"2026-07-05T10:17:03.131585Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:03.131585Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MuDAF: Long-Context Multi-Document Attention Focusing through Contrastive Learning on Attention Heads","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongmei Zhang, Ming Gong, Ning Wu, Shining Liang, Shiping Yang, Weihao Liu, Wenbiao Ding","submitted_at":"2025-02-19T18:59:15Z","abstract_excerpt":"Large Language Models (LLMs) frequently show distracted attention due to irrelevant information in the input, which severely impairs their long-context capabilities. Inspired by recent studies on the effectiveness of retrieval heads in long-context factutality, we aim at addressing this distraction issue through improving such retrieval heads directly. We propose Multi-Document Attention Focusing (MuDAF), a novel method that explicitly optimizes the attention distribution at the head level through contrastive learning. According to the experimental results, MuDAF can significantly improve the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.13963","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/2502.13963/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":"2502.13963","created_at":"2026-07-05T10:17:03.131641+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.13963v1","created_at":"2026-07-05T10:17:03.131641+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.13963","created_at":"2026-07-05T10:17:03.131641+00:00"},{"alias_kind":"pith_short_12","alias_value":"LRQGXTEMNQ7W","created_at":"2026-07-05T10:17:03.131641+00:00"},{"alias_kind":"pith_short_16","alias_value":"LRQGXTEMNQ7WSTVT","created_at":"2026-07-05T10:17:03.131641+00:00"},{"alias_kind":"pith_short_8","alias_value":"LRQGXTEM","created_at":"2026-07-05T10:17:03.131641+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/LRQGXTEMNQ7WSTVTGYYMFB4P23","json":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23.json","graph_json":"https://pith.science/api/pith-number/LRQGXTEMNQ7WSTVTGYYMFB4P23/graph.json","events_json":"https://pith.science/api/pith-number/LRQGXTEMNQ7WSTVTGYYMFB4P23/events.json","paper":"https://pith.science/paper/LRQGXTEM"},"agent_actions":{"view_html":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23","download_json":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23.json","view_paper":"https://pith.science/paper/LRQGXTEM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.13963&json=true","fetch_graph":"https://pith.science/api/pith-number/LRQGXTEMNQ7WSTVTGYYMFB4P23/graph.json","fetch_events":"https://pith.science/api/pith-number/LRQGXTEMNQ7WSTVTGYYMFB4P23/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23/action/storage_attestation","attest_author":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23/action/author_attestation","sign_citation":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23/action/citation_signature","submit_replication":"https://pith.science/pith/LRQGXTEMNQ7WSTVTGYYMFB4P23/action/replication_record"}},"created_at":"2026-07-05T10:17:03.131641+00:00","updated_at":"2026-07-05T10:17:03.131641+00:00"}