Paper Detail

Learn What's Left, Not What's Mastered: Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization

Yixuan Wang, Yifei Chen, Haichao Zhang, Haozheng Luo, Xander Wu, Jie Ni, Yun Fu, Nuno Vasconcelos, Yijiang Li

huggingface Score 13.8

Published 2026-08-17 · First seen 2026-08-18

General AI

Abstract

Reinforcement learning (RL) with group-relative advantages has become the de facto standard for post-training language model reasoners. However, when optimizing multiple reward objectives, existing methods typically scalarize the reward vector with a fixed weighted sum before group-wise standardization. We show that this design leads to two fundamental problems: rollouts with distinct reward profiles can receive identical advantages, and all objectives are optimized with fixed relative weights regardless of their current level of saturation. As a result, training continues to allocate gradient budget to already-solved objectives instead of focusing on those with greater remaining headroom. We introduce Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization (SA-MRPO), which standardizes each reward objective independently and adaptively discounts its contribution according to a batch-level estimate of objective saturation. This dynamically reallocates optimization effort toward under-optimized objectives while empirically maintaining performance on those that are already well satisfied. We further show that saturation-aware reweighting can reverse the sign of an update, rather than merely rescale its magnitude. Across mathematical reasoning with two- and three-objective reward combinations, SA-MRPO improves the harder correctness objective over GDPO in 12 of 15 benchmark comparisons, with gains of up to 5% on AIME24. On adaptive reasoning it improves accuracy on all five benchmarks, by 3.8% on average and up to 9.2 % on AMC23, and on coding benchmarks it improves pass rate by up to 2.3%, while in all settings maintaining the easier objectives near their already satisfied levels.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@misc{wang2026learn,
  title = {Learn What's Left, Not What's Mastered: Saturation Aware Advantage Reweighting for Multi-Reward Policy Optimization},
  author = {Yixuan Wang and Yifei Chen and Haichao Zhang and Haozheng Luo and Xander Wu and Jie Ni and Yun Fu and Nuno Vasconcelos and Yijiang Li},
  year = {2026},
  abstract = {Reinforcement learning (RL) with group-relative advantages has become the de facto standard for post-training language model reasoners. However, when optimizing multiple reward objectives, existing methods typically scalarize the reward vector with a fixed weighted sum before group-wise standardization. We show that this design leads to two fundamental problems: rollouts with distinct reward profiles can receive identical advantages, and all objectives are optimized with fixed relative weights r},
  url = {https://huggingface.co/papers/2608.16072},
  keywords = {group-relative advantages, multi-reward policy optimization, saturation-aware advantage reweighting, SA-MRPO, objective saturation, gradient budget reallocation, huggingface daily},
  eprint = {2608.16072},
  archiveprefix = {arXiv},
}

Metadata

{}