General AI Assistant Tasks on GAIA (Text, MM, File breakdown)
62.1Accuracy (Text)ReAct + ACE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ReAct + ACEMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 62.1 | 50 | 28.9 | 52.7 | |
| ReAct + ACEMain LLM=GPT-4.1, Framework=ReAct2026.06 | 51.5 | 33.3 | 23.7 | 42.4 | |
| ReActMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 51.5 | 41.7 | 34.2 | 46.1 | |
| ReAct + TruncationMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 51.5 | 29.2 | 28.9 | 43 | |
| ReAct + SummarizationMain LLM=GPT-4.1, Framework=ReAct2026.06 | 49.5 | 29.2 | 21.1 | 40 | |
| ReActMain LLM=GPT-4.1, Framework=ReAct2026.06 | 46.7 | 33.3 | 21.1 | 38.8 | |
| ReAct + SummarizationMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 46.6 | 29.2 | 23.7 | 38.8 | |
| ReAct + TruncationMain LLM=GPT-4.1, Framework=ReAct2026.06 | 43.7 | 33.3 | 21.1 | 37 |