Emotional Reasoning on WebEmo Out-of-domain
50.45AccuracyEMO-R3
Evaluation Results
| Method | Links | |
|---|---|---|
| EMO-R3Backbone=Qwen2.5-VL-3B-Instruct, Roll-out=4, Post-training=true2026.02 | 50.45 | |
| DAPOBackbone=Qwen2.5-VL-3B-Instruct, Roll-out=4, Post-training=true2026.02 | 49.8 | |
| EMO-R3Backbone=Qwen2.5-VL-3B-Instruct, Roll-out=8, Post-training=true2026.02 | 49.7 | |
| GRPOBackbone=Qwen2.5-VL-3B-Instruct, Roll-out=4, Post-training=true2026.02 | 49.5 | |
| GRPOBackbone=Qwen2.5-VL-3B-Instruct, Roll-out=8, Post-training=true2026.02 | 49.4 | |
| DAPOBackbone=Qwen2.5-VL-3B-Instruct, Roll-out=8, Post-training=true2026.02 | 48.8 | |
| SEPM*Backbone=LLaVA1.5-7B, Roll-out=-, Post-training=false2026.02 | 42.39 | |
| Vanilla*Backbone=Qwen2.5-VL-3B-Instruct, Roll-out=-, Post-training=false2026.02 | 40.65 | |
| Vanilla*Backbone=LLaVA1.5-7B, Roll-out=-, Post-training=false2026.02 | 25.56 | |
| SFTBackbone=Qwen2.5-VL-3B-Instruct, Roll-out=-, Post-training=true2026.02 | 17.75 |