Sports Understanding on FitnessQA
88.2Accuracydual-pathway self-consistency framework
Evaluation results
| Method | Links | |
|---|---|---|
| dual-pathway self-consistency frameworkBase Model=PerceptionLM-8B, Evaluation Protocol=Finetuned2026.03 | 88.2 | |
| PerceptionLM-8BEvaluation Protocol=Finetuned2026.03 | 74.1 | |
| Qwen3-VL-8B-InstructAnswer Extraction Failure Handling=discarding answer extraction failures2026.03 | 55.4 | |
| Gemini-3-Flash2026.03 | 54.2 | |
| PerceptionLM-8B2026.03 | 39.5 | |
| Qwen2.5-VL-7B-InstructAnswer Extraction Failure Handling=discarding answer extraction failures2026.03 | 33.3 |