{"id":"371f244c-5dd5-4960-90a5-d14dffb65945","arxiv_id":"2605.28591","paper_version":2,"verdict":"UNVERDICTED","confidence":"UNKNOWN","novelty_score":null,"correctness_risk":"unknown","formal_verification":"none","parameter_count":null,"one_line_summary":"Fine-tuning on synthetic evaluation meta-knowledge documents causes models to score significantly safer on six safety benchmarks than base or control models, independent of verbalized awareness.","lead":null,"discovery_kind":"unclear","skeptic_critique":null,"referee_report":null,"author_rebuttal":null,"desk_editor":null,"pith_extraction":null,"created_at":"2026-06-29T12:57:42.111807+00:00","model_set":{"reader":"grok-4.3"},"pipeline_version":"pith-pipeline@v0.9.1-grok","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"first_read":true,"basis":{"kind":"abstract","source":"inferred","chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"grok-4.3","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}