LLM Arbitration on PAVE Dimension 1 Counterfactual Setting v1 (test) benchmark leaderboard · SOTA2 Research

LLM Arbitration on PAVE Dimension 1 Counterfactual Setting v1 (test)

0.661MarginLlama3-8B

Evaluation Results

Method	Links
Llama3-8B2026.05		0.661	29.95	32.49	2.339	46.38	33	9.96	2.03	14.87	61.25	18.8	30.8	8.23	0.445	26.73	40.16	4.83	0.672	12.02	38.75	25.69
GPT-4o-mini2026.05		0.521	40.23	36.9	1.486	61.73	25.33	1.33	2.948	1.78	63.51	25.28	24.73	8.39	0.329	33.92	57.59	1.48	0.345	2.57	36.49	26.62
Deepseek-v32026.05		0.235	71.7	17.19	0.395	60.75	44.05	1.91	1.27	3.42	64.17	45.06	40.75	13.1	0.687	32.15	66.99	2.47	2.03	3.68	35.83	20.26
Mistral-7B2026.05		0.19	71.59	1.55	0.397	5.45	55.76	4.11	0.793	9.29	14.74	9.08	26.36	0.18	0.358	0.68	27.26	23.06	0.375	84.58	85.26	62.02
Qwen3-32B2026.05		0.115	50.59	41.79	0.977	84.57	60.82	1.23	0.644	3.15	87.72	44.7	29.75	3.23	0.423	10.85	47.37	0.68	0.9	1.43	12.28	8.37
Gemini-2.52026.05		0.085	56.33	21.83	0.775	50	48.5	5.15	1.062	10	60	33.02	43.65	8.37	0.776	19.17	60.02	12.5	1.502	20.83	40	19.13
Phi-42026.05		0.004	49.43	11.44	1.023	22.62	48.84	10.08	1.048	19.71	42.33	20.81	42.76	8.26	0.747	19.31	58.57	22.47	1.414	38.36	57.67	26.95