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

Policy Selection on Policy Selection Benchmark

94nDCG@1

GEARS

Updated 5mo ago

Evaluation Results

Method	Links
GEARS 2026.02		94	96	96	94	60	37	0.82	56	94	95	86	94
GEARS 2026.02		87	91	91	87	57	35	0.72	53	89	90	77	87
Code-as-Action 2026.02		77	87	87	77	52	33	0.59	45	84	88	68	77
CoT 2026.02		68	80	83	68	44	30	0.31	43	73	82	57	68
Self-Refine 2026.02		61	78	80	61	44	31	0.34	37	74	83	50	61
Naive 2026.02		57	70	74	57	39	28	0.2	36	67	77	44	57
Self-Consistency 2026.02		57	74	76	57	39	28	0.13	33	67	78	37	57
GEARS 2026.02		40	42	42	40	18	11	0.8	24	32	32	26	40