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

Ethical Reasoning on ETHICS (test)

86Accuracy

MultiRole-R1

Updated 1mo ago

Evaluation Results

Method	Links
MultiRole-R1 2025.07		86	92.89
SelfConsis SFT+GRPO 2025.07		83.5	89.42
MultiRole-R1 2025.07		81.95	89.09
SelfConsis SFT+DPO 2025.07		81.88	42.32
SelfConsis SFT 2025.07		81.45	71.34
SelfConsis SFT+GRPO 2025.07		79.82	86.84
More think 2025.07		79.36	83.33
SelfConsis SFT 2025.07		78.39	68.35
More think 2025.07		78.26	72.45
SelfConsis SFT+DPO 2025.07		78.09	50.31
Role-Play 2025.07		77.16	75.3
Self-Refine 2025.07		76.48	83.22
MultiRole-R1 2025.07		75.84	96.54
SelfConsis SFT+GRPO 2025.07		75.63	87.89
Zero-shot CoT 2025.07		73.63	72.2
Zero-shot CoT 2025.07		72.29	76.68
SelfConsis SFT+DPO 2025.07		71.22	83.64
Role-Play 2025.07		70.67	47.83
Self-Refine 2025.07		70.64	49.6
SelfConsis SFT 2025.07		70.56	81.36
SelfConsis SFT+GRPO 2025.07		69.5	85.4
More think 2025.07		68.06	83.99
SelfConsis SFT 2025.07		67.45	82.19
SelfConsis SFT+DPO 2025.07		67.28	68.51
MultiRole-R1 2025.07		66.83	87.27
More think 2025.07		64.44	81.53
Role-Play 2025.07		64.24	79.78
Zero-shot CoT 2025.07		62.46	79.44
Self-Refine 2025.07		60.96	80.17
Self-Refine 2025.07		52.19	37.36
Zero-shot CoT 2025.07		51.82	36.14
Role-Play 2025.07		50.83	37.89