chinese-ai-detector-bert / eval_comparison.json
AnxForever's picture
Upgrade to v11c boundary-fix (three-set avg 98.56%, independent eval 98.57%, validation 98.75%)
dfd7f4b verified
Raw
History Blame Contribute Delete
2.49 kB
{
"bert_v11c_boundary_fix": {
"model": "bert_v11c_boundary_fix",
"core_v1_test_clean": {
"accuracy": 97.98,
"precision": 97.87,
"recall": 98.77,
"f1": 98.32,
"samples": 545,
"time_sec": 6.2
},
"independent_data": {
"accuracy": 98.57,
"precision": 93.08,
"recall": 98.67,
"f1": 95.79,
"samples": 910,
"time_sec": 9.3
},
"independent_data_by_source": {
"Toutiao_News": {
"accuracy": 100.0,
"count": 221
},
"Toutiao_news_edu": {
"accuracy": 100.0,
"count": 38
},
"Toutiao_news_finance": {
"accuracy": 100.0,
"count": 49
},
"Toutiao_news_tech": {
"accuracy": 100.0,
"count": 69
},
"Wikipedia_CN": {
"accuracy": 99.16,
"count": 119
},
"external_m4_qazh": {
"accuracy": 95.92,
"count": 49
},
"formal_collected": {
"accuracy": 96.5,
"count": 200
},
"real_ai_deepseek-v3.2": {
"accuracy": 100.0,
"count": 8
},
"real_ai_gemini-3-flash-preview": {
"accuracy": 100.0,
"count": 16
},
"real_ai_gemini-3-pro-preview": {
"accuracy": 100.0,
"count": 24
},
"real_ai_gemini-3-pro-preview-search": {
"accuracy": 87.5,
"count": 8
},
"real_ai_glm-4.7": {
"accuracy": 100.0,
"count": 9
},
"real_ai_gpt-4": {
"accuracy": 100.0,
"count": 10
},
"real_ai_gpt-5": {
"accuracy": 100.0,
"count": 8
},
"real_ai_gpt-oss-120b": {
"accuracy": 100.0,
"count": 8
},
"real_ai_llama-3.1-405b-instruct": {
"accuracy": 100.0,
"count": 9
},
"real_ai_m4_chatgpt": {
"accuracy": 98.0,
"count": 50
}
},
"independent_data_errors": 13,
"independent_data_calibration": {
"optimal_T": 0.7872,
"ECE_before": 0.0168,
"ECE_after": 0.0034,
"high_conf_errors_before": 11,
"high_conf_errors_after": 12
},
"merged_v2_val_clean": {
"accuracy": 99.13,
"precision": 98.07,
"recall": 100.0,
"f1": 99.03,
"samples": 1144,
"time_sec": 12.8
},
"three_set_avg": 98.56
}
}