LLM Hallucination Detection on ContextualJudgeBench benchmark leaderboard · SOTA2 Research

LLM Hallucination Detection on ContextualJudgeBench

92Refusal Rate (Ans.)DeepSeek-R1

Evaluation Results

Method	Links
DeepSeek-R12026.03		92	52	72	50.4	41.2	60.6	20.4	26.2	51.9
R1Distill-Llama-3.3-70B2026.03		89.6	50.4	74	48.4	42.4	57.4	19.2	29.5	51.4
SFRJudge-70B2026.03		87.6	32.4	60.8	54.8	40.8	53.4	44.7	36.1	51.4
Skywork-70B2026.03		82.4	11.2	48	47.6	36.8	41.4	21.6	27.9	39.6
Alta-Selene-8B2026.03		74.4	26.4	40.8	32.8	32.4	34.7	23.1	32	37.1
Llama-3.3-70B2026.03		71.6	42.4	68	48.4	42	51.8	20.8	30.7	47
GPT-4o-mini2026.03		71.2	22.8	45.6	42.4	33.2	54.2	11.8	29.5	38.8
SFRJudge-8B2026.03		70.8	22	40.4	38.8	40.4	43.4	27.5	31.1	39.3
SFRJudge-12B2026.03		68.4	28.4	45.2	43.6	28	51	16.1	29.5	38.8
Llama-3-OffsetBias-8B2026.03		64.8	11.2	34	26.4	33.2	21.1	46.3	23	32.6
GPT-4o2026.03		64	52	68	50.8	39.6	56.2	12.9	22.5	45.8
Skywork-8B2026.03		60.8	12	38.8	31.6	38.4	26.7	29.4	21.3	32.4
Llama-3.1-70B2026.03		59.6	48	58	48.4	38	51.8	15.7	27.5	43.4
ST-Eval-70B2026.03		50	42	51.2	45.6	40.8	39.4	36.1	29.9	41.9
MARCH-STEM2026.03		38.1	64	53.6	58	44.9	65.5	44.7	49.5	52.3
MARCH-General2026.03		37.2	57.8	52.6	54.8	48	63.9	48.1	50.1	51.6
Llama-3.1-8B2026.03		28	43.2	34.8	34.8	23.2	41	11.4	21.3	29.7
Promtheus-2-8x7b2026.03		22	29.6	22.4	29.6	20.4	39.8	10.2	18.4	24.1
Promtheus-2-7b2026.03		12.4	44	27.2	32	24	42.6	6.7	29.5	27.3
Glider-3.8B2026.03		12	8.8	45.6	9.2	20.8	28.7	5.1	4.1	16.8