Decision-Oriented LM Jev Improves ICL Example Retrieval, Distilled into Lightweight MLP

A new paper proposes Jev, a decision-oriented language model that directly evaluates the utility of candidate examples for a given query to improve in-context learning example retrieval. The model can be distilled into a lightweight MLP to reduce overhead and speed up retrieval.

2026-11-04 ~ 2026-11-04 · 2 related posts

• Distilling decision-oriented LLM judgments into an MLP speeds up ICL demo retrieval — _reachsumit · 2026-11-04
• Jev model improves ICL demonstration retrieval via distillation into a lightweight MLP — kalyan_kpl · 2026-11-04