High-level Reasoning on HLE (Text, MM, All)
29.7HLE Score (Text)ReAct
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ReActMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 29.7 | 26.5 | 29 | |
| ReAct + ACEMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 29.5 | 29.2 | 29.4 | |
| ReAct + TruncationMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 29.2 | 28.3 | 29 | |
| ReAct + SummarizationMain LLM=Gemini-3.1-flash-lite-preview, Framework=ReAct2026.06 | 25.8 | 26.5 | 26 | |
| ReAct + ACEMain LLM=GPT-4.1, Framework=ReAct2026.06 | 24.8 | 22.1 | 24.2 | |
| ReAct + TruncationMain LLM=GPT-4.1, Framework=ReAct2026.06 | 23.5 | 23 | 23.4 | |
| ReAct + SummarizationMain LLM=GPT-4.1, Framework=ReAct2026.06 | 23 | 23 | 23 | |
| ReActMain LLM=GPT-4.1, Framework=ReAct2026.06 | 22.5 | 21.2 | 22.2 |