{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:E7YU67LUB45OVBAGIIV2JR24XO","short_pith_number":"pith:E7YU67LU","schema_version":"1.0","canonical_sha256":"27f14f7d740f3aea8406422ba4c75cbb99f1fdf6b7e84d0e805f31af924bef25","source":{"kind":"arxiv","id":"2506.11108","version":1},"attestation_state":"computed","paper":{"title":"History-Aware Cross-Attention Reinforcement: Self-Supervised Multi Turn and Chain-of-Thought Fine-Tuning with vLLM","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andreas Lemos, Andrew Kiruluta, Priscilla Burity","submitted_at":"2025-06-08T01:58:17Z","abstract_excerpt":"We present CAGSR-vLLM-MTC, an extension of our Self-Supervised Cross-Attention-Guided Reinforcement (CAGSR) framework, now implemented on the high-performance vLLM runtime, to address both multi-turn dialogue and chain-of-thought reasoning. Building upon our original single-turn approach, we first instrumented vLLM's C++/CUDA kernels to asynchronously capture per-layer, per-head cross-attention weights during generation. We then generalized our self-supervised reward function to accumulate attention signals over entire conversation histories and intermediate chain-of-thought steps. We discuss "},"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":"2506.11108","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-08T01:58:17Z","cross_cats_sorted":[],"title_canon_sha256":"9d01570c91de128b44a711ba6bb960cf7d9676cec0dba36dfa73b2c419d8459f","abstract_canon_sha256":"20e722a796de460ac0ea0319676644a43723a72f15c3cf87330c53ccbce76c06"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:59.564365Z","signature_b64":"8CpMUbXJaY9CSO9jwWRVQ682M8JkFgQchjUaxOjXZE9wkAexAxPQGFW1nYKHEVmFz3HwsDIUGSqH9ZE7KBIjDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"27f14f7d740f3aea8406422ba4c75cbb99f1fdf6b7e84d0e805f31af924bef25","last_reissued_at":"2026-07-05T11:20:59.563917Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:59.563917Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"History-Aware Cross-Attention Reinforcement: Self-Supervised Multi Turn and Chain-of-Thought Fine-Tuning with vLLM","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andreas Lemos, Andrew Kiruluta, Priscilla Burity","submitted_at":"2025-06-08T01:58:17Z","abstract_excerpt":"We present CAGSR-vLLM-MTC, an extension of our Self-Supervised Cross-Attention-Guided Reinforcement (CAGSR) framework, now implemented on the high-performance vLLM runtime, to address both multi-turn dialogue and chain-of-thought reasoning. Building upon our original single-turn approach, we first instrumented vLLM's C++/CUDA kernels to asynchronously capture per-layer, per-head cross-attention weights during generation. We then generalized our self-supervised reward function to accumulate attention signals over entire conversation histories and intermediate chain-of-thought steps. We discuss "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.11108","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/2506.11108/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":"2506.11108","created_at":"2026-07-05T11:20:59.563976+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.11108v1","created_at":"2026-07-05T11:20:59.563976+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.11108","created_at":"2026-07-05T11:20:59.563976+00:00"},{"alias_kind":"pith_short_12","alias_value":"E7YU67LUB45O","created_at":"2026-07-05T11:20:59.563976+00:00"},{"alias_kind":"pith_short_16","alias_value":"E7YU67LUB45OVBAG","created_at":"2026-07-05T11:20:59.563976+00:00"},{"alias_kind":"pith_short_8","alias_value":"E7YU67LU","created_at":"2026-07-05T11:20:59.563976+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/E7YU67LUB45OVBAGIIV2JR24XO","json":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO.json","graph_json":"https://pith.science/api/pith-number/E7YU67LUB45OVBAGIIV2JR24XO/graph.json","events_json":"https://pith.science/api/pith-number/E7YU67LUB45OVBAGIIV2JR24XO/events.json","paper":"https://pith.science/paper/E7YU67LU"},"agent_actions":{"view_html":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO","download_json":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO.json","view_paper":"https://pith.science/paper/E7YU67LU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.11108&json=true","fetch_graph":"https://pith.science/api/pith-number/E7YU67LUB45OVBAGIIV2JR24XO/graph.json","fetch_events":"https://pith.science/api/pith-number/E7YU67LUB45OVBAGIIV2JR24XO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO/action/storage_attestation","attest_author":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO/action/author_attestation","sign_citation":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO/action/citation_signature","submit_replication":"https://pith.science/pith/E7YU67LUB45OVBAGIIV2JR24XO/action/replication_record"}},"created_at":"2026-07-05T11:20:59.563976+00:00","updated_at":"2026-07-05T11:20:59.563976+00:00"}