method · first seen 14 Oct 2026
inverse probability scaling
Also: IPS
Reward reweighting that divides each outcome's reward by its estimated sampling probability to encourage diverse outcomes.
Mentions per week
PapersCodeDiscussion
Earliest sources
- MIPS: Mitigating Mode Collapse in Reinforcement Learning for Large Outcome SpacesPapers · 14 Oct 2026 · Armaan Sandhu, Abhijeet Sinha, Dianbo Liu
Latest sources
- MIPS: Mitigating Mode Collapse in Reinforcement Learning for Large Outcome SpacesPapers · 14 Oct 2026 · Armaan Sandhu, Abhijeet Sinha, Dianbo Liu