{ "model": "/root/xt/work/replay2/merged", "acc": { "arc": 0.8, "mmlu": 0.6375, "hellaswag": 0.4125 }, "cap_score": 0.6166666666666667, "ppl": 31.617075032896558, "ppl_worst_chunk": 45.51285588597565, "ppl_code": NaN, "ppl_code_worst_chunk": NaN, "refusal_propensity": 0.42529962956905376, "refusal_margin_harmful": 1.0696644097566606, "refusal_margin_benign": 0.6443647801876068, "n_refusal_prompts": 20, "n_benign_prompts": 20, "n_mc": 240 }