{"public_id":"cl_9367e0669af0cc39c2ac38d4683abd1a","status":"active","superseded_by_public_id":null,"corpus_id":280291954,"text":"Data augmentation with truncated model outputs as adversarial negative examples produces Master Reward Models with state-of-the-art robustness against master key attacks while preserving standard evaluation performance.","confidence":0.96,"paper":{"corpus_id":280291954,"title":"One Token to Fool LLM-as-a-Judge","url":"https://sah.borca.ai/papers/280291954"},"contributors":[{"id":1,"public_id":"12632b8b5f","public_label":"Anonymous (12632b8b5f)","roles":["extraction"],"url":"https://sah.borca.ai/u/12632b8b5f"}],"origin_summary":{"object_type":"claim","status":"active","confidence":0.96,"origin_kinds":["extraction_create"],"contribution_count":1,"contribution_task_types":["extraction"],"contribution_statuses":["applied"],"verifier_verdict_count":0,"verifier_classes":[],"verifier_class_counts":{"system":0,"user_agent":0},"verdict_counts":{"approve":0,"reject":0},"verifier_state":"no_verdicts","basis":["kg_settlement_results.decision_payload.legacy_bridge","kg_entity_origin_refs","kg_assertion_proposals","contributions","verifications","claim.status","claim.confidence"],"limits":["ledger provenance is aggregated; raw contribution and verifier audit rows are not expanded","entity matching uses settlement bridge refs and edge commands"]},"concepts":[{"public_id":"co_45f3645a0ee759cf714f7dd927b3859d","name":"truncated model outputs","description":"Partially cut off model responses used as adversarial negative examples for training.","types":["training data"],"url":"https://sah.borca.ai/concepts/co_45f3645a0ee759cf714f7dd927b3859d"},{"public_id":"co_87add40ff07144196662c5594b3c17d0","name":"adversarial negative examples","description":"Training examples designed to represent undesirable outputs and improve robustness.","types":["training data"],"url":"https://sah.borca.ai/concepts/co_87add40ff07144196662c5594b3c17d0"},{"public_id":"co_ba45cee954fb80ee76aea31715e246cb","name":"Master Reward Models","description":"Reward models trained with adversarial augmentation to resist superficial attack patterns.","types":["model"],"url":"https://sah.borca.ai/concepts/co_ba45cee954fb80ee76aea31715e246cb"}],"related_claims":[],"url":"https://sah.borca.ai/claims/cl_9367e0669af0cc39c2ac38d4683abd1a"}