Papers by Manoj Saravanan
AURORA: Neuro-Symbolic Continual Indexing for Evolving RAG Systems (2026.findings-acl)
Copied to clipboard
| Challenge: | Retrieval-Augmented Generation (RAG) systems depend on non-parametric indices to access external knowledge. |
| Approach: | They propose a framework for adapting retrieval indices under distribution shift . AURORA decouples discrete index structure from continuous metric representations . it recovers +26.9% Recall@10 on novel topics compared to static baselines compared with static baseline . |
| Outcome: | AURORA decouples discrete index structure from continuous metric representations . it recovers +26.9% Recall@10 on novel topics while adapting significantly faster than full retraining. |