Visual Question Answering on BLINK single-image subset
41.2AccuracyAttWarp-Chains
Evaluation results
| Method | Links | |
|---|---|---|
| AttWarp-ChainsBase MLLM=LLaVA2025.10 | 41.2 | |
| AttWarpBase MLLM=LLaVA2025.10 | 40.4 | |
| AttWarp-DistillBase MLLM=LLaVA2025.10 | 39.7 | |
| Base MLLMBase MLLM=LLaVA2025.10 | 38.3 |