Goal reconstruction on OneStop (New Item)
0.651BERTScoreIncorrect Human (same critical span)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Incorrect Human (same critical span)Human baseline condition=same critical span2025.05 | 0.651 | 67.7 | |
| Gemini few-shotModel backbone=Gemini 3, Evaluation protocol=few-shot2025.05 | 0.642 | 68.3 | |
| DalEye-LlamaInput modality=Text + Eye-movements, Model backbone=Llama 3.12025.05 | 0.631 | 64.8 | |
| DalEye-GPTInput modality=Text + Eye-movements, Model backbone=GPT-4o-mini2025.05 | 0.63 | 65.8 | |
| Gemini zero-shotModel backbone=Gemini 3, Evaluation protocol=zero-shot2025.05 | 0.629 | 66.4 | |
| Text-only GPT-4o-miniInput modality=Text-only, Model backbone=GPT-4o-mini2025.05 | 0.619 | 61.9 | |
| DalEye-LLaVAInput modality=Text + Eye-movements, Model backbone=LLaVA OneVision2025.05 | 0.618 | 61 | |
| Text-only Llama 3.1Input modality=Text-only, Model backbone=Llama 3.12025.05 | 0.617 | 60.9 | |
| Arbitrary Gemini 3Model backbone=Gemini 3, Evaluation protocol=Arbitrary2025.05 | 0.612 | 63.6 | |
| Incorrect Human (different critical span)Human baseline condition=different critical span2025.05 | 0.603 | 49 | |
| Text-only LLaVA OneVisionInput modality=Text-only, Model backbone=LLaVA OneVision2025.05 | 0.595 | 63.6 |