$ python3 -I tools/experiments/charactereval_anchor_diagnostics.py /results/anchor /results/anchor_diagnostics.json # started 2026-10-29T12:00:30+00:00 ended 2026-10-29T12:00:30+00:00 exit 0 D1 rows: {'ChatGLM3-6B': {'rows': 8032, 'duplicates': 0, 'missing': 0, 'extra': 0}, 'Baichuan2-7B': {'rows': 8032, 'duplicates': 0, 'missing': 0, 'extra': 0}, 'XVERSE-7B': {'rows': 8032, 'duplicates': 0, 'missing': 0, 'extra': 0}, 'InternLM-7B': {'rows': 8032, 'duplicates': 0, 'missing': 0, 'extra': 0}, 'Qwen-7B': {'rows': 8032, 'duplicates': 0, 'missing': 0, 'extra': 0}} D2 ChatGLM3-6B: ka:n718 d-0.0073 z-0.35 cs:n734 d-0.0044 z-0.15 coh:n708 d+0.0011 z+0.04 ed:n737 d+0.0043 z+0.14 flu:n308 d+0.0120 z+0.30 ke:n714 d+0.0045 z+0.17 pu:n724 d+0.0007 z+0.03 cons:n715 d-0.0047 z-0.13 pb:n720 d+0.0066 z+0.15 kh:n486 d+0.0086 z+0.32 emp:n744 d-0.0064 z-0.29 hl:n724 d+0.0067 z+0.18 D2 Baichuan2-7B: ka:n718 d-0.0087 z-0.42 cs:n734 d+0.0006 z+0.02 coh:n708 d-0.0016 z-0.07 ed:n737 d-0.0007 z-0.02 flu:n308 d-0.0056 z-0.14 ke:n714 d+0.0038 z+0.15 pu:n724 d-0.0022 z-0.09 cons:n715 d+0.0008 z+0.02 pb:n720 d-0.0099 z-0.24 kh:n486 d+0.0007 z+0.03 emp:n744 d+0.0000 z+0.00 hl:n724 d-0.0032 z-0.09 D2 XVERSE-7B: ka:n718 d+0.0063 z+0.29 cs:n734 d-0.0030 z-0.10 coh:n708 d-0.0022 z-0.09 ed:n737 d-0.0018 z-0.06 flu:n308 d+0.0101 z+0.25 ke:n714 d-0.0042 z-0.16 pu:n724 d+0.0001 z+0.00 cons:n715 d+0.0001 z+0.00 pb:n720 d+0.0054 z+0.13 kh:n486 d-0.0038 z-0.14 emp:n744 d+0.0005 z+0.02 hl:n724 d-0.0041 z-0.11 D2 InternLM-7B: ka:n718 d+0.0016 z+0.08 cs:n734 d+0.0027 z+0.08 coh:n708 d+0.0024 z+0.10 ed:n737 d-0.0002 z-0.01 flu:n308 d+0.0042 z+0.11 ke:n714 d-0.0005 z-0.02 pu:n724 d-0.0029 z-0.12 cons:n715 d+0.0041 z+0.12 pb:n720 d-0.0025 z-0.06 kh:n486 d+0.0063 z+0.23 emp:n744 d+0.0021 z+0.10 hl:n724 d-0.0019 z-0.05 D2 Qwen-7B: ka:n718 d+0.0004 z+0.02 cs:n734 d-0.0053 z-0.17 coh:n708 d-0.0008 z-0.03 ed:n737 d+0.0046 z+0.14 flu:n308 d+0.0019 z+0.04 ke:n714 d+0.0038 z+0.14 pu:n724 d-0.0059 z-0.23 cons:n715 d+0.0119 z+0.33 pb:n720 d+0.0039 z+0.09 kh:n486 d-0.0109 z-0.38 emp:n744 d-0.0008 z-0.04 hl:n724 d-0.0048 z-0.13 D3 implied per-score noise (rms) 0.1211; z mean -0.001, sd 0.166, max |z| 0.423 D4 pairwise orders kept: 150/150 D5 beyond 0.005 (17): [['ChatGLM3-6B', 'flu', 0.012], ['Qwen-7B', 'cons', 0.0119], ['Qwen-7B', 'kh', -0.0109], ['XVERSE-7B', 'flu', 0.0102], ['Baichuan2-7B', 'pb', -0.0099], ['Baichuan2-7B', 'ka', -0.0087], ['ChatGLM3-6B', 'kh', 0.0086], ['ChatGLM3-6B', 'ka', -0.0072], ['ChatGLM3-6B', 'hl', 0.0067], ['ChatGLM3-6B', 'pb', 0.0066], ['ChatGLM3-6B', 'emp', -0.0064], ['XVERSE-7B', 'ka', 0.0063], ['InternLM-7B', 'kh', 0.0063], ['Qwen-7B', 'pu', -0.0059], ['Baichuan2-7B', 'flu', -0.0056], ['XVERSE-7B', 'pb', 0.0054], ['Qwen-7B', 'cs', -0.0054]]