{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:EWOVH3M6VTX6ZIGLLZ53NU45TL","short_pith_number":"pith:EWOVH3M6","canonical_record":{"source":{"id":"2502.10482","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-14T01:44:04Z","cross_cats_sorted":[],"title_canon_sha256":"392f3ab3f8dcb565ed8805b7977f05bc4137f50e6731e0300273d6f71719feaf","abstract_canon_sha256":"3508f0160a1a5add3af70eacbc740ea89919188720143d716a4a83952f6d891a"},"schema_version":"1.0"},"canonical_sha256":"259d53ed9eacefeca0cb5e7bb6d39d9ad678ef59b2fa08e8c51e57c4b3c96f8c","source":{"kind":"arxiv","id":"2502.10482","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.10482","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"arxiv_version","alias_value":"2502.10482v2","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.10482","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"pith_short_12","alias_value":"EWOVH3M6VTX6","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"pith_short_16","alias_value":"EWOVH3M6VTX6ZIGL","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"pith_short_8","alias_value":"EWOVH3M6","created_at":"2026-07-05T10:50:11Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:EWOVH3M6VTX6ZIGLLZ53NU45TL","target":"record","payload":{"canonical_record":{"source":{"id":"2502.10482","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-14T01:44:04Z","cross_cats_sorted":[],"title_canon_sha256":"392f3ab3f8dcb565ed8805b7977f05bc4137f50e6731e0300273d6f71719feaf","abstract_canon_sha256":"3508f0160a1a5add3af70eacbc740ea89919188720143d716a4a83952f6d891a"},"schema_version":"1.0"},"canonical_sha256":"259d53ed9eacefeca0cb5e7bb6d39d9ad678ef59b2fa08e8c51e57c4b3c96f8c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:11.754236Z","signature_b64":"uxzYRbsI1gGbwKHa/bE5b9MDZIyU8SpCy2Jq4J9kJrXltBA4wWeiWVkEL4XKwMIdw+n0DQvnIQ/M8N+cSOovAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"259d53ed9eacefeca0cb5e7bb6d39d9ad678ef59b2fa08e8c51e57c4b3c96f8c","last_reissued_at":"2026-07-05T10:50:11.753796Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:11.753796Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.10482","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:50:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ADz26AW8iPJhBoaAEdPOYu6v+LodIJ6ov2HmVkn/HH2g9QuJ/vwkkzbeV+vlm1jteAtenM9HViZBrCPkq7YBDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T13:05:34.637884Z"},"content_sha256":"b31c0ac5c36ac9eb92704a7d638665e25322efd396dc7092ed16d7eb794a124b","schema_version":"1.0","event_id":"sha256:b31c0ac5c36ac9eb92704a7d638665e25322efd396dc7092ed16d7eb794a124b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:EWOVH3M6VTX6ZIGLLZ53NU45TL","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"A Self-Supervised Reinforcement Learning Approach for Fine-Tuning Large Language Models Using Cross-Attention Signals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Andreas Lemos, Andrew Kiruluta, Priscilla Burity","submitted_at":"2025-02-14T01:44:04Z","abstract_excerpt":"We propose a novel reinforcement learning framework for post training large language models that does not rely on human in the loop feedback. Instead, our approach uses cross attention signals within the model itself to derive a self supervised reward, thereby guiding iterative fine tuning of the model policy. By analyzing how the model attends to the input prompt during generation, we construct measures of prompt coverage, focus, and coherence. We then use these measures to rank or score candidate responses, providing a reward signal that encourages the model to produce well aligned, on topic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.10482","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/2502.10482/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:50:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"PhTK9OC2UX5f4nQp2zuMM2UQiiA1/aPg6ppnBsjy9JKBqPiXdyZi3+G660W1h+VSn0hqslEXUFgwpLQLCgg3AA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T13:05:34.638756Z"},"content_sha256":"ac937e1c8cf178055c42c45f89b09be44ba379f90bd14cc41f85fd72e77505fa","schema_version":"1.0","event_id":"sha256:ac937e1c8cf178055c42c45f89b09be44ba379f90bd14cc41f85fd72e77505fa"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL/bundle.json","state_url":"https://pith.science/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-09T13:05:34Z","links":{"resolver":"https://pith.science/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL","bundle":"https://pith.science/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL/bundle.json","state":"https://pith.science/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL/state.json","well_known_bundle":"https://pith.science/.well-known/pith/EWOVH3M6VTX6ZIGLLZ53NU45TL/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:EWOVH3M6VTX6ZIGLLZ53NU45TL","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"3508f0160a1a5add3af70eacbc740ea89919188720143d716a4a83952f6d891a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-14T01:44:04Z","title_canon_sha256":"392f3ab3f8dcb565ed8805b7977f05bc4137f50e6731e0300273d6f71719feaf"},"schema_version":"1.0","source":{"id":"2502.10482","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.10482","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"arxiv_version","alias_value":"2502.10482v2","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.10482","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"pith_short_12","alias_value":"EWOVH3M6VTX6","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"pith_short_16","alias_value":"EWOVH3M6VTX6ZIGL","created_at":"2026-07-05T10:50:11Z"},{"alias_kind":"pith_short_8","alias_value":"EWOVH3M6","created_at":"2026-07-05T10:50:11Z"}],"graph_snapshots":[{"event_id":"sha256:ac937e1c8cf178055c42c45f89b09be44ba379f90bd14cc41f85fd72e77505fa","target":"graph","created_at":"2026-07-05T10:50:11Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2502.10482/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We propose a novel reinforcement learning framework for post training large language models that does not rely on human in the loop feedback. Instead, our approach uses cross attention signals within the model itself to derive a self supervised reward, thereby guiding iterative fine tuning of the model policy. By analyzing how the model attends to the input prompt during generation, we construct measures of prompt coverage, focus, and coherence. We then use these measures to rank or score candidate responses, providing a reward signal that encourages the model to produce well aligned, on topic","authors_text":"Andreas Lemos, Andrew Kiruluta, Priscilla Burity","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-14T01:44:04Z","title":"A Self-Supervised Reinforcement Learning Approach for Fine-Tuning Large Language Models Using Cross-Attention Signals"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.10482","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b31c0ac5c36ac9eb92704a7d638665e25322efd396dc7092ed16d7eb794a124b","target":"record","created_at":"2026-07-05T10:50:11Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"3508f0160a1a5add3af70eacbc740ea89919188720143d716a4a83952f6d891a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-14T01:44:04Z","title_canon_sha256":"392f3ab3f8dcb565ed8805b7977f05bc4137f50e6731e0300273d6f71719feaf"},"schema_version":"1.0","source":{"id":"2502.10482","kind":"arxiv","version":2}},"canonical_sha256":"259d53ed9eacefeca0cb5e7bb6d39d9ad678ef59b2fa08e8c51e57c4b3c96f8c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"259d53ed9eacefeca0cb5e7bb6d39d9ad678ef59b2fa08e8c51e57c4b3c96f8c","first_computed_at":"2026-07-05T10:50:11.753796Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:50:11.753796Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"uxzYRbsI1gGbwKHa/bE5b9MDZIyU8SpCy2Jq4J9kJrXltBA4wWeiWVkEL4XKwMIdw+n0DQvnIQ/M8N+cSOovAg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:50:11.754236Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.10482","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b31c0ac5c36ac9eb92704a7d638665e25322efd396dc7092ed16d7eb794a124b","sha256:ac937e1c8cf178055c42c45f89b09be44ba379f90bd14cc41f85fd72e77505fa"],"state_sha256":"35bf428cefd6ddfc7cabe420daea94cc002df440bcf8d02a084506efb3f5bb6c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1K53Cqe0S/HpG6b4FjySTW2XPQ6gZyZ9kyRwsStpnBdVAErjILmz3T60bMZGmHgsarj6HtyK5RibVJ/EOEWICA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T13:05:34.646665Z","bundle_sha256":"c23b32b276449cc2a27668420327423a369f7be930c7666a98bcda26535690ef"}}