VQA on RexTime 1.0 (test)
0.791AccuracyVITAL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VITALParam. (B)=7B, Type=Reinforcement Learning2026.02 | 0.791 | — | |
| TimeSearchParam. (B)=7B, Type=Reinforcement Learning2026.02 | 0.765 | 0.2944 | |
| GPT-4oParam. (B)=-, Type=Zero-shot2026.02 | 0.7367 | 0.2867 | |
| GraphThinkerParam. (B)=7B, Type=Reinforcement Learning2026.02 | 0.713 | 0.3075 | |
| Claude3-OpusParam. (B)=-, Type=Zero-shot2026.02 | 0.6867 | 0.1367 | |
| Gemini-1.5-ProParam. (B)=-, Type=Zero-shot2026.02 | 0.68 | 0.1833 | |
| GraphThinker(w/o RL)Param. (B)=7B, Type=Zero-shot2026.02 | 0.6682 | 0.1521 | |
| GPT-4VParam. (B)=-, Type=Zero-shot2026.02 | 0.6333 | 0.1667 | |
| Reka-CoreParam. (B)=-, Type=Zero-shot2026.02 | 0.5967 | 0.17 | |
| VTimeLLMParam. (B)=7B, Type=Supervised Fine-Tuning2026.02 | 0.5758 | 0.1713 | |
| Qwen2.5-VLParam. (B)=7B, Type=Zero-shot2026.02 | 0.566 | 0.0635 | |
| TimeChatParam. (B)=7B, Type=Supervised Fine-Tuning2026.02 | 0.4946 | 0.1092 | |
| TimeChatParam. (B)=7B, Type=Zero-shot2026.02 | 0.4004 | — | |
| VTimeLLMParam. (B)=7B, Type=Zero-shot2026.02 | 0.3616 | — | |
| LITAParam. (B)=13B, Type=Zero-shot2026.02 | 0.3444 | — |