NPO paper: prompt optimization may hinge on teacher feedback, not search trees
rohanpaul_ai · x · 2026-09-02
- A paper compares NPO with GEPA for prompt optimization: GEPA maintains multiple prompt candidates with Pareto-based selection, while NPO keeps a single prompt lineage and asks a teacher model to revise it using recent rollout traces and rewards.
- Result: NPO matches or beats GEPA, suggesting prompt optimization may not need an elaborate search tree at all—the quality of the teacher and its feedback matters more.
- Counterintuitive takeaway for prompt engineers: invest in the revision feedback loop rather than complex search structures.
More from Research
- TRACE Open-Sourced: Migrating G1 Controller to Full-Size Humanoid H2 in 8 Hours on One GPU — zhengyiluo · 2026-10-18
- Humanoid Robot Funding Tops $3B as Brian Potter Explains the VLA Models Driving Robots — scottleibrand · 2026-10-18
- Terence Tao on 'War Footing' as OpenAI Claims Navier-Stokes Solve Shakes Mathematics — AlexTensor · 2026-10-18
- Univ. of Tokyo paper: why your recorded voice feels eerie — a vocal uncanny valley — ssh4net · 2026-10-18
- MIT's SANDO Drone Planner Avoids Every Moving Obstacle Across 12 Real Flights — Olivier__OG · 2026-10-18
- No-backprop pretraining speedup of 2.3x claimed, inspired by mech interp essay — CatAstro_Piyush · 2026-10-18