RAG-Diff Uses Retrieval to Steer Frozen Policies
Nekovowo · x · 2026-07-13
RAG-Diff proposes a novel framework that uses retrieval at test time to steer frozen policies, aiming to help pre-trained policies better adapt to individual preferences in tasks like caregiving.
The author notes that while pre-trained policies can handle some caregiving tasks, real-world care requires "person-specific adjustments":
- The core method is test-time steering: instead of modifying the original policy parameters, it uses RAG for dynamic guidance during inference.
- This framework targets scenarios where individual preferences frequently change, focusing on making frozen policies more adaptable.
Related event: RAG-Diff Framework Adapts Robot Caregiving to Human Preferences(4 posts)→
More from Research
- Researcher flags unresolved data transparency issues in compositionality evals — najoungkim · 2026-10-27
- Separate Forms from Estimators: Why Steer by Difference of Means When You Can Use SGD? — johnhewtt · 2026-10-27
- EMNLP paper probes why some LLM biases survive rewording while others don't — radamihalcea · 2026-10-27
- Datology webinar: Curation Studio matched top data baselines with under 20% compute — josh_wills · 2026-10-27
- KANLab: an interactive browser playground for Kolmogorov-Arnold Networks — Lukas Niehaus · 2026-10-27
- CMU's LLM2Cert synthesizes stability certificates verified via SMT and Lean — CarnegieMellonU · 2026-10-27