Research map / Continual learning / Online and replay-based
RAG without Forgetting: Continual Query-Infused Key Memory
Yuntong Hu, Sha Li, Naren Ramakrishnan, Liang Zhao
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 ERM key evolution | BM25 · BEIR+BRIGHT average (13 datasets) | nDCG@1: 26.3 -> 38.5 | Improved |
| add ERM key evolution | BGE-Large · BEIR+BRIGHT average (13 datasets) | nDCG@1: 48.6 -> 55.7 | Improved |
| add ERM key evolution | GTE-Base · BEIR+BRIGHT average (13 datasets) | nDCG@1: 49.9 -> 56.4 | Improved |
| add ERM key evolution | Cohere · BEIR+BRIGHT average (13 datasets) | nDCG@1: 48.7 -> 55.2 | Improved |
| add ERM key evolution | Voyage · BEIR+BRIGHT average (13 datasets) | nDCG@1: 50.8 -> 56.3 | Improved |
| add ERM key evolution | BM25 · BRIGHT AoPS | nDCG@1: 0.9 -> 20.7 | Improved |
| add ERM key evolution | BGE-Large · BRIGHT Pony | nDCG@1: 43.1 -> 87.5 | Improved |
| add ERM key evolution | BGE-Large · BRIGHT Biology | nDCG@1: 95.1 -> 91.3 | Hurt |
| add ERM key evolution | BGE-Large · BRIGHT Robotics | nDCG@1: 43.4 -> 40.4 | Hurt |
| add ERM key evolution | dense retrievers · BRIGHT StackExchange label-disjoint queries | nDCG@1: within ±3% of baseline | No clear effect |
| add ERM key evolution | BM25 · BRIGHT StackExchange QA | answer quality: 72.6 -> 76.6 | Improved |
| add ERM key evolution | BGE-Large · BRIGHT StackExchange QA | answer quality: 74.5 -> 77.6 | Improved |
Closest papers by meaning
- Adapting to Non-Stationary Environments: Multi-Armed Bandit Enhanced Retrieval-Augmented Generation on Knowledge Graphs (AAAI 2025 · similarity 0.89)
- RAGGED: Towards Informed Design of Scalable and Stable RAG Systems (ICML 2025 · similarity 0.88)
- DynamicRAG: Leveraging Outputs of Large Language Model as Feedback for Dynamic Reranking in Retrieval-Augmented Generation (NeurIPS 2025 · similarity 0.88)
- Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting (ICLR 2025 · similarity 0.88)
- From RAG to Memory: Non-Parametric Continual Learning for Large Language Models (ICML 2025 · similarity 0.87)
- Long-Context LLMs Meet RAG: Overcoming Challenges for Long Inputs in RAG (ICLR 2025 · similarity 0.87)
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.
