Code-generation monitoring on Evaluation objectives Standard prompt setting (out-of-domain) benchmark leaderboard · SOTA2 Research

Code-generation monitoring on Evaluation objectives Standard prompt setting (out-of-domain)

100Exit RateGemini-3-Flash

Evaluation Results

Method	Links
Gemini-3-Flash2026.05		100	100	100	21.82	78.57	100	100	100	100	88.93
Gemini-3-Flash2026.05		100	100	100	3.27	37.17	100	97.3	100	100	81.97
Gemini-3-Flash2026.05		100	100	100	16.36	90.48	100	100	100	100	89.65
Gemini-3-Flash2026.05		94.12	100	100	3.33	15.04	100	95.95	93.33	100	77.97
CoT-Guard 4B2026.05		88.89	79.17	88.64	34.18	42.77	80	88.68	81.93	90	74.92
CoT-Guard 4B2026.05		86.27	87.5	88.64	24.73	28.82	100	94.3	95.26	83.33	76.54
Qwen3-32B2026.05		85.29	100	97.73	3.27	1.77	93.75	30.41	17.33	19.61	49.91
Qwen3-4B2026.05		85.19	70.83	90.91	27.82	37.8	80	97.3	83.68	90	73.73
Qwen3-4B2026.05		82.68	91.67	90.91	25.21	31.23	100	90.33	38.62	93.33	71.55
Qwen3-32B2026.05		82.14	100	100	6.67	1.77	100	81.1	76	62.22	67.77
GPT-5.42026.05		73.5	100	100	0	0	100	56.8	56	30.43	57.41
Qwen3-8B2026.05		65.14	95.83	88.67	21.64	6.74	90	39.2	58.67	75.75	60.18
Qwen3-8B2026.05		62.31	95.83	97.73	0	6.19	100	12.16	1.29	26.09	44.62
Qwen3-4B2026.05		53.81	95.83	93.91	0	4.42	100	27.03	6.44	8.7	43.35
Qwen3-4B2026.05		50	91.67	96.15	6.42	3.37	100	60.8	60	51.4	57.76
GPT-5.42026.05		38.24	100	100	0	0	93.75	0	1.33	0	37.04
GPT-5-mini2026.05		26.47	95.45	80.65	0	1.77	81.25	25.68	54.67	24.64	43.4
Llama3.1-8B-Instruct2026.05		22.88	20	12.87	9.45	3.24	60	27.2	9.2	22.32	20.8
GPT-5-mini2026.05		20.59	100	50	0	0	81.25	1.35	0	0	28.13
Qwen3-4B2026.05		16.99	95.83	50	0	0	93.5	9.8	20	28.02	34.9
Llama3.1-8B-Instruct2026.05		13.07	10	3.52	2.42	0.78	3.12	7.98	6.44	12.75	6.68
Qwen3-4B2026.05		8.82	86.84	20.97	0	0	75	0	1.33	8.7	22.41