ReMax achieves the first sublinear regret bound for Gaussian rewards at M=2 by characterizing the optimal sampling distribution via an expected-improvement balance condition and separating saturation from underestimation effects.
Optimizing language models for inference time objectives using reinforcement learning.arXiv preprint arXiv:2503.19595
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 4verdicts
UNVERDICTED 4roles
background 1polarities
background 1representative citing papers
Tail-extrapolated estimators approximate best-of-N policy gradients from limited training rollouts by leveraging upper-tail reward statistics under structural assumptions.
Derives new loss functions for SFT and RL that optimize directly for test-time inference operators like aggregation or filtering, with empirical gains in scaling.
Introduces polychromic objectives adapted into PPO via vine sampling and modified advantages, showing higher success rates and better coverage under perturbations on BabyAI, Minigrid, and algorithmic tasks.
citing papers explorer
-
Finite-Time Regret Analysis of Retry-Aware Bandits
ReMax achieves the first sublinear regret bound for Gaussian rewards at M=2 by characterizing the optimal sampling distribution via an expected-improvement balance condition and separating saturation from underestimation effects.
-
What should post-training optimize? A test-time scaling law perspective
Tail-extrapolated estimators approximate best-of-N policy gradients from limited training rollouts by leveraging upper-tail reward statistics under structural assumptions.
-
Compute Aligned Training: Optimizing for Test Time Inference
Derives new loss functions for SFT and RL that optimize directly for test-time inference operators like aggregation or filtering, with empirical gains in scaling.