{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WI67VKJHARZ26PBW5TKG62MVCG","short_pith_number":"pith:WI67VKJH","schema_version":"1.0","canonical_sha256":"b23dfaa9270473af3c36ecd46f69951182e38987fbb0a5842ec4226751008acb","source":{"kind":"arxiv","id":"2512.17375","version":2},"attestation_state":"computed","paper":{"title":"AdvJudge-Zero: Binary Decision Flips in LLM-as-a-Judge via Adversarial Control Tokens","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Hongliang Liu, Tung-Ling Li, Yuhao Wu","submitted_at":"2025-12-19T09:22:11Z","abstract_excerpt":"LLM-as-a-Judge systems supply the reward signal in modern RLHF and RLVR pipelines, but their binary verdict reduces to a single linear readout F_gap on one hidden state. We show this readout is shallow enough that short, low-perplexity tokens flip the verdict from \"No\" to \"Yes\". These tokens are sampled from the judge's own next-token distribution at the response position, with no manual seed set and no gradient-based optimization. Our procedure, AdvJudge-Zero, reaches $>$90% ensemble false-positive rate on 22 of 24 (model, dataset) cells across six Qwen, Llama, and Gemma judges, versus 54-72%"},"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":"2512.17375","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-12-19T09:22:11Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"f45a4de46355cc3ae2cfc8defeb32085472bf5aba97429012b810a8a900cd443","abstract_canon_sha256":"2b337e48ffb43864b346fd5d29457429612f02d6a93dafdcdb2362170bf49dbf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:07.659915Z","signature_b64":"yoqI+FwMvOBzhnCBlsgPF0LlmKdA9OP9vXtU+IXfZpShnX42m3yxmtg+1r6ymhYUg5uyrI/U3S5bj6MaFGrQAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b23dfaa9270473af3c36ecd46f69951182e38987fbb0a5842ec4226751008acb","last_reissued_at":"2026-05-28T01:04:07.659226Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:07.659226Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AdvJudge-Zero: Binary Decision Flips in LLM-as-a-Judge via Adversarial Control Tokens","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Hongliang Liu, Tung-Ling Li, Yuhao Wu","submitted_at":"2025-12-19T09:22:11Z","abstract_excerpt":"LLM-as-a-Judge systems supply the reward signal in modern RLHF and RLVR pipelines, but their binary verdict reduces to a single linear readout F_gap on one hidden state. We show this readout is shallow enough that short, low-perplexity tokens flip the verdict from \"No\" to \"Yes\". These tokens are sampled from the judge's own next-token distribution at the response position, with no manual seed set and no gradient-based optimization. Our procedure, AdvJudge-Zero, reaches $>$90% ensemble false-positive rate on 22 of 24 (model, dataset) cells across six Qwen, Llama, and Gemma judges, versus 54-72%"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2512.17375","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/2512.17375/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":"2512.17375","created_at":"2026-05-28T01:04:07.659307+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.17375v2","created_at":"2026-05-28T01:04:07.659307+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.17375","created_at":"2026-05-28T01:04:07.659307+00:00"},{"alias_kind":"pith_short_12","alias_value":"WI67VKJHARZ2","created_at":"2026-05-28T01:04:07.659307+00:00"},{"alias_kind":"pith_short_16","alias_value":"WI67VKJHARZ26PBW","created_at":"2026-05-28T01:04:07.659307+00:00"},{"alias_kind":"pith_short_8","alias_value":"WI67VKJH","created_at":"2026-05-28T01:04:07.659307+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/WI67VKJHARZ26PBW5TKG62MVCG","json":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG.json","graph_json":"https://pith.science/api/pith-number/WI67VKJHARZ26PBW5TKG62MVCG/graph.json","events_json":"https://pith.science/api/pith-number/WI67VKJHARZ26PBW5TKG62MVCG/events.json","paper":"https://pith.science/paper/WI67VKJH"},"agent_actions":{"view_html":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG","download_json":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG.json","view_paper":"https://pith.science/paper/WI67VKJH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.17375&json=true","fetch_graph":"https://pith.science/api/pith-number/WI67VKJHARZ26PBW5TKG62MVCG/graph.json","fetch_events":"https://pith.science/api/pith-number/WI67VKJHARZ26PBW5TKG62MVCG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG/action/storage_attestation","attest_author":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG/action/author_attestation","sign_citation":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG/action/citation_signature","submit_replication":"https://pith.science/pith/WI67VKJHARZ26PBW5TKG62MVCG/action/replication_record"}},"created_at":"2026-05-28T01:04:07.659307+00:00","updated_at":"2026-05-28T01:04:07.659307+00:00"}