Atomic Action Recognition on SceneTeract (test)
77.1AccuracyGemini-3-Flash-Preview
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-3-Flash-PreviewEvaluation Setting=Decomposed2026.03 | 77.1 | |
| Qwen3-VL-4B-Instruct with GRPOEvaluation Setting=Decomposed2026.03 | 75.3 | |
| Gemma3-12B-InstructEvaluation Setting=Decomposed2026.03 | 75 | |
| Gemma3-4B-InstructEvaluation Setting=Decomposed2026.03 | 74.8 | |
| Claude-Sonnet-4-6Evaluation Setting=Decomposed2026.03 | 73.1 | |
| Gemini-3.1-Pro-PreviewEvaluation Setting=Decomposed2026.03 | 72.1 | |
| Qwen3-VL-4B-InstructEvaluation Setting=Decomposed2026.03 | 70.4 | |
| Qwen3-VL-8B-InstructEvaluation Setting=Decomposed2026.03 | 67.5 | |
| Ministral3-3B-InstructEvaluation Setting=Decomposed2026.03 | 34.5 |