{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QH7YROHDI2KMERXY4PYAQHNO55","short_pith_number":"pith:QH7YROHD","schema_version":"1.0","canonical_sha256":"81ff88b8e34694c246f8e3f0081daeef5fe2efa4d845dcff0e40a84d584490f7","source":{"kind":"arxiv","id":"2605.19377","version":1},"attestation_state":"computed","paper":{"title":"The Evaluation Game: Beyond Static LLM Benchmarking","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anne-Marie Kermarrec, Jade Garcia-Bourr\\'ee, Paul Wang, Vincent Corruble","submitted_at":"2026-05-19T05:22:21Z","abstract_excerpt":"As jailbreaks, adversarially crafted inputs that bypass safety constraints, continue to be discovered in Large Language Models, practitioners increasingly rely on fine-tuning as a defensive strategy. Yet the theoretical foundations underlying this robustness fine-tuning remain underexplored. We introduce a game-theoretic framework in which the interaction between an evaluator (auditing the model for jailbreaks) and a trainer is formalized as a two-player game.\n  A key feature of our approach is the use of group actions, a mathematical structure that captures symmetries and transformations, to "},"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":"2605.19377","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-19T05:22:21Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"791a64563c30c11b64e7f92290ffdd42e1b9676b2f69028b4f8ab1c60858fdec","abstract_canon_sha256":"b6f65e463bca20fe0863be78ebdfc87bf6ea164b79f3dc0b0e861e5c86b2bbd6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:42.627597Z","signature_b64":"DggC+3UvEiqPx+6IsfbYW8Z8bZj1FiMTEy5JIa9OWH8SYFWLIOZABFXBSLINBq1Shu0IjOHhMoQupl1bzFObCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"81ff88b8e34694c246f8e3f0081daeef5fe2efa4d845dcff0e40a84d584490f7","last_reissued_at":"2026-05-20T01:05:42.626912Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:42.626912Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Evaluation Game: Beyond Static LLM Benchmarking","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anne-Marie Kermarrec, Jade Garcia-Bourr\\'ee, Paul Wang, Vincent Corruble","submitted_at":"2026-05-19T05:22:21Z","abstract_excerpt":"As jailbreaks, adversarially crafted inputs that bypass safety constraints, continue to be discovered in Large Language Models, practitioners increasingly rely on fine-tuning as a defensive strategy. Yet the theoretical foundations underlying this robustness fine-tuning remain underexplored. We introduce a game-theoretic framework in which the interaction between an evaluator (auditing the model for jailbreaks) and a trainer is formalized as a two-player game.\n  A key feature of our approach is the use of group actions, a mathematical structure that captures symmetries and transformations, to "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19377","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/2605.19377/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":"2605.19377","created_at":"2026-05-20T01:05:42.627019+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19377v1","created_at":"2026-05-20T01:05:42.627019+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19377","created_at":"2026-05-20T01:05:42.627019+00:00"},{"alias_kind":"pith_short_12","alias_value":"QH7YROHDI2KM","created_at":"2026-05-20T01:05:42.627019+00:00"},{"alias_kind":"pith_short_16","alias_value":"QH7YROHDI2KMERXY","created_at":"2026-05-20T01:05:42.627019+00:00"},{"alias_kind":"pith_short_8","alias_value":"QH7YROHD","created_at":"2026-05-20T01:05:42.627019+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/QH7YROHDI2KMERXY4PYAQHNO55","json":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55.json","graph_json":"https://pith.science/api/pith-number/QH7YROHDI2KMERXY4PYAQHNO55/graph.json","events_json":"https://pith.science/api/pith-number/QH7YROHDI2KMERXY4PYAQHNO55/events.json","paper":"https://pith.science/paper/QH7YROHD"},"agent_actions":{"view_html":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55","download_json":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55.json","view_paper":"https://pith.science/paper/QH7YROHD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19377&json=true","fetch_graph":"https://pith.science/api/pith-number/QH7YROHDI2KMERXY4PYAQHNO55/graph.json","fetch_events":"https://pith.science/api/pith-number/QH7YROHDI2KMERXY4PYAQHNO55/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55/action/storage_attestation","attest_author":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55/action/author_attestation","sign_citation":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55/action/citation_signature","submit_replication":"https://pith.science/pith/QH7YROHDI2KMERXY4PYAQHNO55/action/replication_record"}},"created_at":"2026-05-20T01:05:42.627019+00:00","updated_at":"2026-05-20T01:05:42.627019+00:00"}