LLM Negotiation on In-distribution (held-out) benchmark leaderboard · SOTA2 Research

LLM Negotiation on In-distribution (held-out)

0.58RewardQwen3-30B-A3B-Instruct-2507-trained (Ours)

Evaluation Results

Method
Qwen3-30B-A3B-Instruct-2507-trained (Ours)2026.07	0.58	70	82.8	2.47	99.5	0
Llama-3.3-70B-Instruct2026.07	0.468	16.2	36.3	2.53	48.4	35.9
Qwen3-30B-A3B-nothink2026.07	0.386	5.5	16	2.22	51.2	38.3
Qwen3-4B-Instruct-25072026.07	0.27	8	9.8	2.35	52	21.9
Qwen3-30B-A3B-Instruct-2507-untrained2026.07	0.151	4.8	35.5	2.39	57.1	14.8
gpt-5.1-high-reasoning2026.07	0.123	18.1	68	2.94	38.2	0
Kimi-K2.6-nothink2026.07	0.101	12	49.2	2.54	39.7	7.4
Qwen3-235B-A22B-Instruct-25072026.07	0.099	11.6	56.2	1.96	44.4	5.9
Kimi-K2.6-thinking2026.07	0.077	12.8	63.8	2.2	48.4	0.4
gpt-5.1-no-reasoning2026.07	0.046	11.5	65.3	1.74	42.9	0.4
DeepSeek-V3.1-nothink2026.07	0.041	19.2	53.9	2.38	35.5	3.9
Qwen3-30B-A3B-think2026.07	0.021	6.2	35.9	2.46	44.6	3.9
DeepSeek-V3.1-thinking2026.07	0.016	14.8	50.8	2.62	34.6	3.1