Complex Reasoning on BIG-bench Hard (Subtask Breakdown) benchmark leaderboard · SOTA2 Research

Complex Reasoning on BIG-bench Hard (Subtask Breakdown)

88Abstract Algebra ScoreInfiGFusion

Evaluation Results

Method	Links
InfiGFusion2025.05		88	95.3	90.91	73.97	54.22	74.6	81.22	87.73	88.4	96.8	96.8	97.6	94	89.2	96.4	99.6	40	70.05	84.16
Phi4 (SFT)2025.05		86	92.74	90.91	74.75	53.61	78.57	79.59	87.73	88	90	95.6	96.4	92.4	88.8	96.8	62	13.2	35.83	77.94
DeepSeek-R1-Distill-Qwen2025.05		85	89.32	82.64	73.41	54.22	91.27	78.37	85.28	39.6	80.8	82.8	83.2	86	83.2	87.6	81.6	9.6	45.35	73.29
Phi42025.05		82	92.74	91.74	75.75	53.01	77.78	77.14	86.5	88.8	94.4	96.8	98.4	85.6	88.4	96.4	64	11.2	32.99	77.43
Qwen2.5-Instruct2025.05		73	90.17	81.82	71.96	52.41	68.25	77.96	85.28	82.8	80.4	79.2	97.6	80.4	68.8	93.6	96.4	35.6	43.85	75.53
Mistral2025.05		65	92.31	81.82	68.6	49.4	66.67	78.78	84.66	76.8	96.4	99.2	98.8	82	62.8	90	93.2	37.2	68.98	77.37