Share your thoughts, 1 month free Claude Pro on usSee more

Bargaining on AmazonHistoryPrice (held-out test)

0.7664Reward

Qwen3-30B-A3B-Instruct-2507-trained

Updated 5d ago

Evaluation Results

Method	Links
Qwen3-30B-A3B-Instruct-2507-trained 2026.04		0.7664	91.99	83.85	0.1
gpt-5.4-high-reasoning 2026.04		0.4021	91.8	43.8	0
gpt-5.4-no-reasoning 2026.04		0.354	84.38	42.85	0.78
Kimi-K2-Thinking 2026.04		0.3419	91.21	36.99	0.2
DeepSeek-V3.1-thinking 2026.04		0.324	91.8	32.66	0
gpt-oss-120b-noreason 2026.04		0.2857	78.12	40.29	0.39
DeepSeek-V3.1-nothink 2026.04		0.2605	85.94	32.89	1.56
gpt-oss-120b-reason 2026.04		0.2337	77.54	34.9	0.2
Qwen3-30B-A3B-nothink 2026.04		0.2177	52.93	4.99	25.39
gpt-oss-20b-reason 2026.04		0.1168	66.99	17.15	0.78
Qwen3-4B-Instruct-2507 2026.04		0.113	51.56	9.01	6.45
Qwen3-235B-A22B-Instruct-2507 2026.04		0.1115	79.3	16.81	2.34
gpt-oss-20b-noreason 2026.04		0.0626	58.4	12.65	9.77
Qwen3-30B-A3B-Instruct-2507-untrained 2026.04		0.0126	73.44	23.92	10.35
Qwen3-30B-A3B-think 2026.04		0.0069	66.8	9.48	6.25
Llama-3.3-70B-Instruct 2026.04		0.0068	74.8	15.8	15.62