{ "hacked_run_dir": "/root/workspace/Shaer/grpo/outputs/train/shaer_grpo_20260411_223409", "hacked_step": 3100, "train_manifest_path": "/root/workspace/Shaer/grpo/outputs/curated_meter_count_easyfirst_short_drop_trio/cap_1000/selected_manifest.csv", "prompt_file": "/root/workspace/Shaer/grpo/prompts/judge_quality_v4.yaml", "golden_summary": { "count": 30, "reward_total": { "n": 30, "mean": 0.8445466952739816, "median": 0.9227769097980243, "min": 0.5078271677918281, "p10": 0.6306678980962714, "p25": 0.8113067406413482, "p75": 0.9410595720407352, "p90": 0.9476963660542256, "max": 0.9654430799169427 }, "meter": { "n": 30, "mean": 0.9576655101608114, "median": 0.9898881312815053, "min": 0.5395800608267627, "p10": 0.848249732191195, "p25": 0.9805628310292759, "p75": 0.9948783055038647, "p90": 0.9973563416849837, "max": 0.9981962209330739 }, "count_adherence": { "n": 30, "mean": 1.0, "median": 1.0, "min": 1.0, "p10": 1.0, "p25": 1.0, "p75": 1.0, "p90": 1.0, "max": 1.0 }, "judge_quality": { "n": 30, "mean": 0.7906666666666669, "median": 0.92, "min": 0.45, "p10": 0.45, "p25": 0.72, "p75": 0.92, "p90": 0.923, "max": 0.95 }, "repeat_soft": { "n": 30, "mean": 0.9838626053754798, "median": 0.9943360071301247, "min": 0.8383064516129032, "p10": 0.9625, "p25": 0.9832476452251293, "p75": 1.0, "p90": 1.0, "max": 1.0 }, "hard_gate_pass_rate": 1.0, "exact_count_rate": 1.0, "latin_leakage_rate": 0.0, "digit_leakage_rate": 0.0, "artifact_leakage_rate": 0.0 }, "hacked_summary": { "count": 39, "reward_total": { "n": 39, "mean": 0.0, "median": 0.0, "min": 0.0, "p10": 0.0, "p25": 0.0, "p75": 0.0, "p90": 0.0, "max": 0.0 }, "meter": { "n": 39, "mean": 0.8468841133626961, "median": 0.9950388649607371, "min": 0.015073049191457575, "p10": 0.0780608298449938, "p25": 0.9434806730369525, "p75": 0.9991970501141816, "p90": 0.9993906755198334, "max": 0.9994455539632088 }, "count_adherence": { "n": 39, "mean": 0.9937728937728937, "median": 1.0, "min": 0.8571428571428572, "p10": 1.0, "p25": 1.0, "p75": 1.0, "p90": 1.0, "max": 1.0 }, "judge_quality": { "n": 39, "mean": 0.16282051282051285, "median": 0.15, "min": 0.15, "p10": 0.15, "p25": 0.15, "p75": 0.15, "p90": 0.25, "max": 0.25 }, "repeat_soft": { "n": 39, "mean": 0.7031907759133498, "median": 0.7136784083192772, "min": 0.4325178355458288, "p10": 0.5164881378882488, "p25": 0.6008119084274752, "p75": 0.8067508194727036, "p90": 0.8878927664770718, "max": 0.9079545454545456 }, "hard_gate_pass_rate": 0.0, "exact_count_rate": 0.9487179487179487, "latin_leakage_rate": 0.0, "digit_leakage_rate": 1.0, "artifact_leakage_rate": 0.0 }, "health": { "healthy": true, "checks": { "golden_total_gt_hacked_total_by_0.15": true, "golden_judge_gt_hacked_judge_by_0.15": true, "golden_judge_not_too_harsh": true, "golden_meter_not_broken": true, "hacked_total_sufficiently_low": true }, "failed_checks": [] } }