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

Domain-specific Reasoning on LegalBench

85.26Accuracy

Reflexion(3 iters)

Updated 25d ago

Evaluation Results

Method	Links
Reflexion(3 iters) 2026.03		85.26	-
Reflexion(3 iters) 2026.03		81.05	-
RM-Regen 2026.03		78.95	-
RM-Regen 2026.03		75.79	-
Reflexion(3 iters) 2026.03		75.79	-
RM-Primed 2026.03		67.37	66.01
RM-Regen 2026.03		66.32	-
RM-Primed (R+) 2026.03		64.21	64.99
RM-Primed 2026.03		64.21	68.23
Best-of-N (N=3) 2026.03		64.21	-
RM-Primed (R+) 2026.03		62.11	65.94
RM-Regen 2026.03		60	-
Contrastive CoT 2026.03		57.89	59.56
RM-Regen 2026.03		56.84	-
Contrastive CoT 2026.03		55.79	60.33
ST CoT 2026.03		55.79	-
Contrastive CoT 2026.03		53.68	61.13
RM-Regen 2026.03		51.58	-
ReflectEvo 2026.03		50.53	-
Few-shot CoT 2026.03		48.42	64.19
ReflectEvo 2026.03		48.42	-
Self-Refine 2026.03		46.32	-
Best-of-N (N=3) 2026.03		45.26	-
ST CoT 2026.03		45.26	-
Few-shot CoT 2026.03		44.21	61.15
ProCo 2026.03		44.21	-
ProCo 2026.03		41.05	-
Self-Refine 2026.03		41.05	-
ProCo 2026.03		38.95	-
Contrastive CoT 2026.03		35.75	57.01
Best-of-N (N=3) 2026.03		32.63	-
Self-Refine 2026.03		28.42	-
ST CoT 2026.03		23.16	-