Embodied Task Completion on ALFWorld (IOD)
96.67AccuracySKILLGEN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SKILLGENModel Weight Type=Proprietary, Model Name=GPT-5.4-Mini, Skill Augmentation Protocol=SKILL2026.05 | 96.67 | 7.33 | |
| SKILLGENModel Weight Type=Proprietary, Model Name=GPT-5.4-Nano, Skill Augmentation Protocol=SKILL2026.05 | 94 | 11.33 | |
| GPT-5.4-MiniModel Weight Type=Proprietary, Skill Augmentation Protocol=BASE2026.05 | 89.33 | — | |
| SKILLGENModel Weight Type=Open-weight, Model Name=Gemma-4-26B, Skill Augmentation Protocol=SKILL2026.05 | 86 | 7.33 | |
| GPT-5.4-NanoModel Weight Type=Proprietary, Skill Augmentation Protocol=BASE2026.05 | 82.67 | — | |
| Gemma-4-26BModel Weight Type=Open-weight, Skill Augmentation Protocol=BASE2026.05 | 78.67 | — | |
| SKILLGENModel Weight Type=Open-weight, Model Name=Qwen-2.5-7B, Skill Augmentation Protocol=SKILL2026.05 | 77.33 | 13.33 | |
| SKILLGENModel Weight Type=Proprietary, Model Name=Grok-4-Fast, Skill Augmentation Protocol=SKILL2026.05 | 77.33 | 10.67 | |
| SKILLGENModel Weight Type=Open-weight, Model Name=Llama-3.1-8B, Skill Augmentation Protocol=SKILL2026.05 | 70.67 | 2.67 | |
| Llama-3.1-8BModel Weight Type=Open-weight, Skill Augmentation Protocol=BASE2026.05 | 68 | — | |
| SKILLGENModel Weight Type=Proprietary, Model Name=Claude-Haiku-4.5, Skill Augmentation Protocol=SKILL2026.05 | 66.67 | 6.67 | |
| Grok-4-FastModel Weight Type=Proprietary, Skill Augmentation Protocol=BASE2026.05 | 66.67 | — | |
| Qwen-2.5-7BModel Weight Type=Open-weight, Skill Augmentation Protocol=BASE2026.05 | 64 | — | |
| Mistral-NemoModel Weight Type=Open-weight, Skill Augmentation Protocol=BASE2026.05 | 63.33 | — | |
| SKILLGENModel Weight Type=Open-weight, Model Name=Mistral-Nemo, Skill Augmentation Protocol=SKILL2026.05 | 62.67 | -0.67 | |
| Claude-Haiku-4.5Model Weight Type=Proprietary, Skill Augmentation Protocol=BASE2026.05 | 60 | — |