Multi-turn Agentic Task on ScienceWorld
62Success RateReMem
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ReMemLLM Backbone=Claude 3.7 Sonnet2025.11 | 62 | 89 | |
| ReMemLLM Backbone=Gemini 2.5 Flash2025.11 | 58 | 81 | |
| EVOLIB2026.05 | 57.4 | 84.3 | |
| DCType=test-time learning (TTL)2026.05 | 55.9 | 83.6 | |
| BaseType=base sampling2026.05 | 53.3 | 82.9 | |
| ExpRecentLLM Backbone=Gemini 2.5 Flash2025.11 | 53 | 83 | |
| ExpRAGLLM Backbone=Gemini 2.5 Flash2025.11 | 53 | 81 | |
| ExpRAGType=test-time learning (TTL)2026.05 | 51.1 | 80.4 | |
| ExpRecentLLM Backbone=Claude 3.7 Sonnet2025.11 | 49 | 82 | |
| ExpRAGLLM Backbone=Claude 3.7 Sonnet2025.11 | 46 | 76 | |
| ReActLLM Backbone=Claude 3.7 Sonnet2025.11 | 44 | 77 | |
| Mem0LLM Backbone=Claude 3.7 Sonnet2025.11 | 37 | 76 | |
| AmemLLM Backbone=Gemini 2.5 Flash2025.11 | 36 | 74 | |
| ReActLLM Backbone=Gemini 2.5 Flash2025.11 | 34 | 71 | |
| SelfRAGLLM Backbone=Gemini 2.5 Flash2025.11 | 34 | 74 | |
| DC-RSLLM Backbone=Claude 3.7 Sonnet2025.11 | 34 | 74 | |
| AWMLLM Backbone=Claude 3.7 Sonnet2025.11 | 34 | 74 | |
| DC-RSLLM Backbone=Gemini 2.5 Flash2025.11 | 33 | 73 | |
| AWMLLM Backbone=Gemini 2.5 Flash2025.11 | 33 | 73 | |
| AmemLLM Backbone=Claude 3.7 Sonnet2025.11 | 33 | 73 | |
| DC-CuLLM Backbone=Claude 3.7 Sonnet2025.11 | 33 | 75 | |
| Mem0LLM Backbone=Gemini 2.5 Flash2025.11 | 32 | 70 | |
| HistoryLLM Backbone=Claude 3.7 Sonnet2025.11 | 32 | 74 | |
| HistoryLLM Backbone=Gemini 2.5 Flash2025.11 | 31 | 71 | |
| SelfRAGLLM Backbone=Claude 3.7 Sonnet2025.11 | 31 | 74 | |
| DC-CuLLM Backbone=Gemini 2.5 Flash2025.11 | 29 | 71 | |
| BaselineLLM Backbone=Gemini 2.5 Flash2025.11 | 24 | 59 | |
| BaselineLLM Backbone=Claude 3.7 Sonnet2025.11 | 10 | 53 |