Research map / Large language models / Retrieval and knowledge
Look Back to Reason Forward: Revisitable Memory for Long-Context LLM Agents
Yaorui Shi, Yuxin Chen, Siyuan Wang, Sihang Li, Hengxing Cai, Qi GU, Xiang Wang, An Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add history-callback memory retrieval to RL-trained memory agent (ReMemR1 vs MemAgent) | MemAgent (Qwen2.5-7B) · HotpotQA (6400 docs) | accuracy: 75.8 -> 80.8 | Improved |
| add callback retrieval | MemAgent (Qwen2.5-3B) · 2WikiMultiHopQA (50 docs) | accuracy: 41.4 -> 53.5 | Improved |
| add callback retrieval | MemAgent (Qwen2.5-3B) · HotpotQA (50 docs) | accuracy: 70.3 -> 70.9 | Improved |
| add callback retrieval at short context without RL | MemAgent (Qwen2.5-3B, no RL) · HotpotQA (50 docs) | accuracy: 60.2 -> 35.4 | Hurt |
| add callback retrieval at long context without RL | MemAgent (Qwen2.5-3B, no RL) · HotpotQA (6400 docs) | accuracy: 14.1 -> 20.5 | Improved |
| use rule-based question-as-query callback instead of RL-learned callback (ablation) | MemAgent (Qwen2.5-3B) · HotpotQA (400 docs) | accuracy: 68.8 -> 60.9 | Hurt |
| use rule-based question-as-query callback (ablation) | MemAgent (Qwen2.5-3B) · HotpotQA (6400 docs) | accuracy: 58.8 -> 60.9 | Improved |
| use rule-based callback (ablation) | MemAgent (Qwen2.5-3B) · 2WikiMultiHopQA (400 docs) | accuracy: 41.4 -> 35.2 | Hurt |
| add step-level state rewards (alpha 1.0 -> 0.8) (ablation) | ReMemR1 (Qwen2.5-3B) · HotpotQA (400 docs) | accuracy: 59.6 -> 74.0 | Improved |
| lower alpha to 0.2 (overweight step-level reward) (ablation) | ReMemR1 (Qwen2.5-3B) · HotpotQA (6400 docs) | accuracy: 66.1 -> 52.0 | Hurt |
| add RL training (ablation) | ReMemR1 (Qwen2.5-3B) · HotpotQA (6400 docs) | accuracy: 20.5 -> 66.1 | Improved |
| alpha 1.0 vs 0.8 at 100 docs (ablation) | ReMemR1 (Qwen2.5-3B) · HotpotQA (100 docs) | accuracy: 73.4 -> 71.7 | Hurt |
Closest papers by meaning
- MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents (ICLR 2026 · similarity 0.91)
- Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents (ICLR 2026 · similarity 0.91)
- Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactions (ICLR 2026 · similarity 0.91)
- Harness the Memory: A Holistic Evaluation of Memory Substrates in Memory Agents (NeurIPS 2026 · similarity 0.90)
- LoongRL: Reinforcement Learning for Advanced Reasoning over Long Contexts (ICLR 2026 · similarity 0.90)
- PlugMem: A Task-Agnostic Plugin Memory Module for LLM Agents (ICML 2026 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
