{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4I63GVLR4B5YOG3TU36R227RMD","short_pith_number":"pith:4I63GVLR","schema_version":"1.0","canonical_sha256":"e23db35571e07b871b73a6fd1d6bf160e75076063027862cde6f21eed32c238b","source":{"kind":"arxiv","id":"2502.06738","version":1},"attestation_state":"computed","paper":{"title":"Resurrecting saturated LLM benchmarks with adversarial encoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dmitrii Volkov, Igor Ivanov","submitted_at":"2025-02-10T18:07:09Z","abstract_excerpt":"Recent work showed that small changes in benchmark questions can reduce LLMs' reasoning and recall. We explore two such changes: pairing questions and adding more answer options, on three benchmarks: WMDP-bio, GPQA, and MMLU variants. We find that for more capable models, these predictably reduce performance, essentially heightening the performance ceiling of a benchmark and unsaturating it again. We suggest this approach can resurrect old benchmarks."},"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":"2502.06738","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-10T18:07:09Z","cross_cats_sorted":[],"title_canon_sha256":"bb6dbb7b2ec89cdf84309dbde9529d8205867068cb9bc1aae9bc88fe04039a29","abstract_canon_sha256":"b6e6342cd2390c91f873a39127458512c562be49154399bead5e53ddd609a465"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:12:11.523187Z","signature_b64":"r6EsmPZ27xRPfBy9Xhq63IBWRnfrxaFPJFx+qprUP/pyXrmRkQIdAZ5lP8NUPeg3FhJ0ULKjr8GFA0G/u9doAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e23db35571e07b871b73a6fd1d6bf160e75076063027862cde6f21eed32c238b","last_reissued_at":"2026-07-05T10:12:11.522732Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:12:11.522732Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Resurrecting saturated LLM benchmarks with adversarial encoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dmitrii Volkov, Igor Ivanov","submitted_at":"2025-02-10T18:07:09Z","abstract_excerpt":"Recent work showed that small changes in benchmark questions can reduce LLMs' reasoning and recall. We explore two such changes: pairing questions and adding more answer options, on three benchmarks: WMDP-bio, GPQA, and MMLU variants. We find that for more capable models, these predictably reduce performance, essentially heightening the performance ceiling of a benchmark and unsaturating it again. We suggest this approach can resurrect old benchmarks."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.06738","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/2502.06738/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":"2502.06738","created_at":"2026-07-05T10:12:11.522779+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.06738v1","created_at":"2026-07-05T10:12:11.522779+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.06738","created_at":"2026-07-05T10:12:11.522779+00:00"},{"alias_kind":"pith_short_12","alias_value":"4I63GVLR4B5Y","created_at":"2026-07-05T10:12:11.522779+00:00"},{"alias_kind":"pith_short_16","alias_value":"4I63GVLR4B5YOG3T","created_at":"2026-07-05T10:12:11.522779+00:00"},{"alias_kind":"pith_short_8","alias_value":"4I63GVLR","created_at":"2026-07-05T10:12:11.522779+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/4I63GVLR4B5YOG3TU36R227RMD","json":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD.json","graph_json":"https://pith.science/api/pith-number/4I63GVLR4B5YOG3TU36R227RMD/graph.json","events_json":"https://pith.science/api/pith-number/4I63GVLR4B5YOG3TU36R227RMD/events.json","paper":"https://pith.science/paper/4I63GVLR"},"agent_actions":{"view_html":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD","download_json":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD.json","view_paper":"https://pith.science/paper/4I63GVLR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.06738&json=true","fetch_graph":"https://pith.science/api/pith-number/4I63GVLR4B5YOG3TU36R227RMD/graph.json","fetch_events":"https://pith.science/api/pith-number/4I63GVLR4B5YOG3TU36R227RMD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD/action/storage_attestation","attest_author":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD/action/author_attestation","sign_citation":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD/action/citation_signature","submit_replication":"https://pith.science/pith/4I63GVLR4B5YOG3TU36R227RMD/action/replication_record"}},"created_at":"2026-07-05T10:12:11.522779+00:00","updated_at":"2026-07-05T10:12:11.522779+00:00"}