Retrieval-Augmented Question Answering on DeepSearch 2wiki
68Success Rate (SR)Training-Free GRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Training-Free GRPOTaskExecutor=Gemini-3.0-flash2026.04 | 68 | 39.1 | |
| ActiveContextTaskExecutor=Gemini-3.0-flash2026.04 | 68 | 7.5 | |
| Full ContextTaskExecutor=Gemini-3.0-flash2026.04 | 62 | 38.9 | |
| SynapseTaskExecutor=Gemini-3.0-flash2026.04 | 42 | 83 | |
| Mem0TaskExecutor=gpt-4o-mini, Category=Baselines & Retrieval-Based Memory2026.04 | 42 | 46.7 | |
| Gemini-3.0-flashTaskExecutor=gpt-4o-mini, Category=Zero-Shot Curators2026.04 | 42 | 9.5 | |
| Full ContextTaskExecutor=gpt-4o-mini, Category=Baselines & Retrieval-Based Memory2026.04 | 40 | 33.8 | |
| Training-Free GRPOTaskExecutor=gpt-4o-mini, Category=Baselines & Retrieval-Based Memory2026.04 | 40 | 36.1 | |
| AWMTaskExecutor=gpt-4o-mini, Category=Baselines & Retrieval-Based Memory2026.04 | 39 | 34.2 | |
| SynapseTaskExecutor=gpt-4o-mini, Category=Baselines & Retrieval-Based Memory2026.04 | 38 | 45 | |
| ContextCuratorTaskExecutor=gpt-4o-mini, Category=RL-Trained Curator2026.04 | 38 | 7.4 | |
| ActiveContextTaskExecutor=Gemini-2.5-flash2026.04 | 36.4 | 9.6 | |
| Gemini-2.5-flashTaskExecutor=gpt-4o-mini, Category=Zero-Shot Curators2026.04 | 34 | 9.1 | |
| SynapseTaskExecutor=Gemini-2.5-flash2026.04 | 32 | 92.4 | |
| gpt-4oTaskExecutor=gpt-4o-mini, Category=Zero-Shot Curators2026.04 | 32 | 11.2 | |
| gpt-4o-miniTaskExecutor=gpt-4o-mini, Category=Zero-Shot Curators2026.04 | 30 | 10 | |
| Qwen2.5-7B-InstructTaskExecutor=gpt-4o-mini, Category=Zero-Shot Curators2026.04 | 28.6 | 10.5 | |
| Full ContextTaskExecutor=Gemini-2.5-flash2026.04 | 28 | 40.8 | |
| Full ContextTaskExecutor=gpt-4o2026.04 | 28 | 38.2 | |
| SynapseTaskExecutor=gpt-4o2026.04 | 28 | 45.9 | |
| Training-Free GRPOTaskExecutor=gpt-4o2026.04 | 28 | 38.9 | |
| ActiveContextTaskExecutor=gpt-4o2026.04 | 26 | 9.3 | |
| Training-Free GRPOTaskExecutor=Gemini-2.5-flash2026.04 | 24 | 78.3 |