Social Reasoning on EmoBench
80.39AccuracyO3
Evaluation results
| Method | Links | |
|---|---|---|
| O32026.03 | 80.39 | |
| GPT-52026.03 | 80.3 | |
| GPT-5Mode=COT2026.03 | 78.82 | |
| GPT-4oMode=COT2026.03 | 77.62 | |
| GPT-4o2026.03 | 76.94 | |
| Distill-Llama-70B2026.03 | 75.96 | |
| DeepSeek-R12026.03 | 73.87 | |
| O3Mode=COT2026.03 | 72.25 | |
| SocialR1-8BAblation=Full2026.03 | 70.1 | |
| SocialR1-8BAblation=w/o Rstruct2026.03 | 70 | |
| SocialR1-8BAblation=only Rout2026.03 | 69.68 | |
| LLaMa3.1-70BMode=COT2026.03 | 68.91 | |
| SocialR1-8BAblation=w/o Rlen2026.03 | 68.6 | |
| SocialR1-4BAblation=Full2026.03 | 68.04 | |
| SocialR1-8BAblation=w/o Rcont2026.03 | 67.78 | |
| SocialR1-4BAblation=w/o Rstruct2026.03 | 67.74 | |
| Qwen3-32BThinking=Disabled2026.03 | 67 | |
| LLaMa3.1-70B2026.03 | 66.32 | |
| SocialR1-4BAblation=w/o Rlen2026.03 | 65.91 | |
| Qwen3-8BThinking=Disabled2026.03 | 64.84 | |
| SocialR1-4BAblation=only Rout2026.03 | 64.22 | |
| Qwen3-32B2026.03 | 63.56 | |
| SocialR1-4BAblation=w/o Rcont2026.03 | 62.51 | |
| Qwen3-8B2026.03 | 59.2 | |
| Qwen3-4B2026.03 | 53.08 | |
| Qwen3-4BThinking=Disabled2026.03 | 52.8 |