Tag Detection on SenBen 2,000 frames (test)
80.6F1 Score (Tag)Gemini 3 Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini 3 ProReasoning mode=low reas., Zero-shot inference=true2026.04 | 80.6 | |
| Gemini 3 FlashReasoning mode=low reas., Zero-shot inference=true2026.04 | 78.4 | |
| GPT-5-miniReasoning mode=med. reas., Zero-shot inference=true2026.04 | 65.9 | |
| Claude Opus 4.6Zero-shot inference=true2026.04 | 65.8 | |
| Claude Sonnet 4.6Zero-shot inference=true2026.04 | 64.3 | |
| GPT-5.2Reasoning mode=med. reas., Zero-shot inference=true2026.04 | 60.8 | |
| Q2L-balParams=241M2026.04 | 59.4 | |
| GPT-5.2Zero-shot inference=true2026.04 | 55 | |
| GLM-4.6VParams=10.3B, Reasoning mode=reas., Zero-shot inference=true2026.04 | 49.2 | |
| Qwen3-VL-8BParams=8.3B, Zero-shot inference=true2026.04 | 46.9 | |
| Q2L-aggParams=241M2026.04 | 45.7 |