Scientific Question Answering on GPQA Diamond (ACC, TOK, η)

53.03Accuracy (ACC)

Qwen3-4B-Thinking-2507 + DEER

Updated 21d ago

Evaluation Results

Method
Qwen3-4B-Thinking-2507 + DEER 2026.05	53.03	4,762	1.414
Qwen3-4B-Thinking-2507 + BET 2026.05	52.53	3,835	1.739
Qwen3-4B-Thinking-2507 2026.05	51.01	6,475	1
Qwen3-4B-Thinking-2507 + VeriThinker 2026.05	50.1	5,986	1.062
Qwen3-4B-Thinking-2507 + DR.SAF 2026.05	50	4,237	1.498
Qwen3-4B-Thinking-2507 + Length-Penalty 2026.05	49.49	4,705	1.335
Qwen3-4B-Thinking-2507 + Overthink 2026.05	48.99	7,568	0.822
Qwen3-4B-Thinking-2507 + DiffAdapt 2026.05	48.48	5,961	1.032
DeepSeek-R1-Distill-Qwen-14B 2026.05	47.58	4,294	1
DeepSeek-R1-Distill-Qwen-14B + BET 2026.05	46.57	2,023	2.078
DeepSeek-R1-Distill-Qwen-14B + DR.SAF 2026.05	45.86	2,643	1.566
DeepSeek-R1-Distill-Qwen-14B + Length-Penalty 2026.05	45.66	3,891	1.059
Qwen3-4B-Thinking-2507 + ThinkSwitcher 2026.05	45.45	6,420	0.899
DeepSeek-R1-Distill-Qwen-7B + VeriThinker 2026.05	43.33	3,948	1.22
DeepSeek-R1-Distill-Qwen-7B + DEER 2026.05	42.32	3,773	1.247
DeepSeek-R1-Distill-Qwen-7B + Overthink 2026.05	41.82	4,351	1.068
DeepSeek-R1-Distill-Qwen-7B 2026.05	41.31	4,592	1
DeepSeek-R1-Distill-Qwen-7B + DiffAdapt 2026.05	40.3	4,012	1.117
DeepSeek-R1-Distill-Qwen-7B + BET 2026.05	39.8	1,987	2.227
DeepSeek-R1-Distill-Qwen-7B + DR.SAF 2026.05	39.3	2,816	1.551
DeepSeek-R1-Distill-Qwen-7B + Length-Penalty 2026.05	38.28	4,540	0.937
DeepSeek-R1-Distill-Qwen-7B + ThinkSwitcher 2026.05	37.27	3,889	1.065