Character Role-playing on CharacterEval benchmark leaderboard · SOTA2 Research

Character Role-playing on CharacterEval

4.14CoherenceDoubao-1.5-Pro-Character

Evaluation Results

Method	Links
Doubao-1.5-Pro-Character2026.06		4.14	3.97	3.77	2.51	3.19	3.25	3.79	3.33	3.73	3.64	3.23	3.42	3.5
DeepSeek-V4-Flash2026.06		4.09	3.86	3.73	2.62	3.14	3.24	3.78	3.26	3.68	3.75	3.13	3.4	3.47
GPT-5-Chat2026.06		4.08	3.99	3.83	2.5	3.15	3.23	4.05	3.35	3.88	3.25	3.56	3.21	3.5
GPT-4o-mini2026.06		3.93	3.5	3.52	2.66	2.99	3.1	3.34	3.07	3.12	3.68	2.65	3.31	3.23
Qwen2.5-14B-Instruct2026.06		3.89	3.58	3.56	2.35	3.02	3	3.09	3.06	3.44	3.2	2.47	3.18	3.13
Qwen3-4B-Instruct2026.06		3.88	3.6	3.51	2.81	3.15	3.17	4.05	3.19	3.52	3.39	3.75	3.23	3.44
Qwen2.5-7B-Instruct2026.06		3.87	3.56	3.53	2.29	2.93	2.94	2.99	3.01	3.31	3.32	2.39	3.16	3.09
Qwen2.5-7B-Instruct + Psy-CoT + RAPO2026.06		3.87	3.58	3.62	2.55	3.17	3.07	3.81	3.09	3.37	3.35	3.24	3.35	3.33
Qwen2.5-14B-Instruct-AdaMARP2026.06		3.85	3.83	3.52	1.7	2.78	2.83	2.81	3.05	3.74	2.49	2.2	2.83	2.94
Qwen2.5-7B-Instruct + Psy-CoT + DAPO2026.06		3.8	3.44	3.47	2.62	3.14	3.01	3.7	3.05	3.19	3.5	3.06	3.32	3.27
Qwen2.5-7B-Instruct + Psy-CoT + GSPO2026.06		3.77	3.28	3.44	2.81	3.15	3.06	3.64	2.95	2.96	3.69	3.12	3.37	3.27
Qwen2.5-7B-Instruct-CoSER2026.06		3.75	3.71	3.42	1.76	2.8	2.75	2.55	3.01	3.69	2.44	2.03	2.84	2.87
Qwen2.5-7B-Instruct-AdaMARP2026.06		3.74	3.62	3.42	1.75	2.81	2.76	2.52	2.94	3.57	2.5	2.02	2.81	2.84
Qwen2.5-7B-Instruct-Crab2026.06		3.68	3.62	3.42	1.74	2.8	2.76	2.53	2.94	3.54	2.49	2.06	2.83	2.84
Llama-3.1-8B-Instruct2026.06		3.66	3.41	3.31	2.08	2.74	2.76	3.12	2.86	3.17	2.9	2.58	2.87	2.94
Llama-3.1-8B-Crab2026.06		3.52	3.39	3.21	1.64	2.66	2.57	2.32	2.78	3.35	2.28	1.93	2.67	2.67
Llama-3.1-8B-CoSER2026.06		3.47	3.14	3.14	1.79	2.6	2.53	2.65	2.59	3.1	2.49	2.17	2.67	2.68
Qwen2.5-7B-Instruct + Psy-CoT + GRPO2026.06		3.33	3.2	3.62	3.04	3.14	3.37	2.74	3.7	3.06	3.81	3.43	3.09	3.29