Question Answering on HotpotQA (Reward Metric)
0.56RewardERL
Evaluation results
| Method | Links | |
|---|---|---|
| ERLBackbone=Qwen3-4B-Instruct-2507, Method variant=Full2026.02 | 0.56 | |
| ERL w/o Mem.Backbone=Qwen3-4B-Instruct-2507, Method variant=Without memory reuse2026.02 | 0.56 | |
| ERLBackbone=Olmo3-7B-Instruct, Method variant=Full2026.02 | 0.5 | |
| ERL w/o Refl.Backbone=Qwen3-4B-Instruct-2507, Method variant=Without structured reflection2026.02 | 0.48 | |
| RLVRBackbone=Olmo3-7B-Instruct2026.02 | 0.47 | |
| ERL w/o Mem.Backbone=Olmo3-7B-Instruct, Method variant=Without memory reuse2026.02 | 0.47 | |
| ERL w/o Refl.Backbone=Olmo3-7B-Instruct, Method variant=Without structured reflection2026.02 | 0.46 | |
| RLVRBackbone=Qwen3-4B-Instruct-25072026.02 | 0.45 |