Image Captioning on SenBen 2,000 frames (test)
77.2CAP ScoreQ2L-agg
Evaluation Results
| Method | Links | |
|---|---|---|
| Q2L-aggParams=241M2026.04 | 77.2 | |
| Q2L-balParams=241M2026.04 | 77.1 | |
| Gemini 3 FlashReasoning mode=low reas., Zero-shot inference=true2026.04 | 65.4 | |
| Gemini 3 ProReasoning mode=low reas., Zero-shot inference=true2026.04 | 64.2 | |
| GPT-5.2Reasoning mode=med. reas., Zero-shot inference=true2026.04 | 61.6 | |
| GPT-5-miniReasoning mode=med. reas., Zero-shot inference=true2026.04 | 60.5 | |
| Claude Opus 4.6Zero-shot inference=true2026.04 | 59.8 | |
| Claude Sonnet 4.6Zero-shot inference=true2026.04 | 59 | |
| GPT-5.2Zero-shot inference=true2026.04 | 58.3 | |
| GLM-4.6VParams=10.3B, Reasoning mode=reas., Zero-shot inference=true2026.04 | 56.3 | |
| Qwen3-VL-8BParams=8.3B, Zero-shot inference=true2026.04 | 54.8 |