arXiv subjects

Xinmu Ge

Publications and source records attributed to Xinmu Ge.

3 recordsLinked to original sources

arXiv2026

Towards Understanding On-Policy Distillation through the Lens of Test-Time Scaling

On-policy distillation (OPD) has emerged as a promising post-training technique for enhancing LLM reasoning. It is commonly believed to enable the student model to distill knowledge from a stronger teacher model, thereby expanding capabilities beyond the pre-OPD base model. In this study, we examine this view through the lens of test-time scaling by varying the sampling budget K and evaluating performance with pass@K and avg@K. Specifically, across several OPD variants, we observe that OPD-trained models maintain superior avg@K performance across sampling budgets, while the advantage in pass@K gradually shifts to the pre-OPD base models as K increases. These results suggest that OPD primarily improves sampling efficiency rather than consistently expanding the student's reasoning capability boundary. The pass@K dynamics throughout OPD training further reveal a progressive shift toward stronger small-K performance at the expense of the large-K capability boundary. Furthermore, a problem-level solvability analysis using pass@1024 as the criterion reveals an asymmetry: OPD causes more previously solvable problems to become unsolvable than previously unsolvable problems to become solvable. Together, these findings suggest that, from the perspective of capability expansion, OPD behaves more like an "illusory distillation": its apparent gains arise primarily from improved sampling efficiency rather than from acquiring genuinely new reasoning capabilities from the teacher.

cs.LG↗

arXiv2026

Focal Reward: Balanced Reinforcement Learning under Rubric-Based Rewards

The open-ended generation in LLMs usually requires multi-dimensional rubrics to adequately assess quality and guide the improvement of reinforcement learning. However, a critical dilemma inherent in this training paradigm is the imbalanced reward polarization along different rubric dimensions. Under this bottleneck, even if LLMs achieve relatively high rewards after training, they may still exhibit severe deficiencies in certain dimensions, leading to a direct deterioration in user experience. To address this problem, we propose Focal Reward, a novel objective to automatically balance the training of reinforcement learning under rubric-based rewards. Specifically, we first leverage an inverse reward projection mechanism to estimate the saturation degree of each criterion in the rubric, which forms the basis to calibrate the reward direction. Then, the final objective is designed with an automatically reweighting coefficient for each criterion to achieve the fine-grained balancing. Extensive experiments across three model scales and six benchmarks demonstrate that our Focal Reward method outperforms the strongest static aggregation baseline in all 18 model-benchmark comparisons. Rollout, mechanism, and ablation analyses further show that these gains arise from online, saturation-aware reallocation toward rubrics that still have room for improvement.

cs.LG↗

arXiv2025

Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models

While reinforcement learning with verifiable rewards (RLVR) is effective to improve the reasoning ability of large language models (LLMs), its reliance on human-annotated labels leads to the scaling up dilemma, especially for complex tasks. Recent self-rewarding methods investigate a label-free alternative to unlock the reasoning capabilities of LLMs, yet they frequently encounter the non-negligible training collapse issue, as the single-view supervision signal easily forms the self-consistent illusion, yielding the reward hacking. Inspired by the success of self-supervised learning, we propose \textit{Co-rewarding}, a novel self-supervised RL framework that improves training stability by seeking complementary supervision from another views. Specifically, we instantiate Co-rewarding in two ways: (1) \textit{Co-rewarding-I} is a data-side instantiation that derives reward signals from contrastive agreement across semantically analogous questions; and (2) \textit{Co-rewarding-II} is a model-side instantiation that maintains a slowly-updated reference teacher with pseudo labels to realize self-distillation. Intuitively, such instantiations introduce different levels of discrepancy to increase the difficulty of training collapse on trivial reasoning solutions. We also explore their orthogonally combined version to further boost the performance. Empirically, Co-rewarding exhibits stable training across various setups, and outperforms other self-rewarding baselines by $+3.31\%$ improvements on average on multiple mathematical reasoning benchmarks, especially by $+7.49\%$ on Llama-3.2-3B-Instruct. Notably, Co-rewarding reaches or even surpasses RLVR with ground-truth (GT) label in several cases, such as a Pass@$1$ of $94.01\%$ on GSM8K with Qwen3-8B-Base remarkably higher than GT. Our code is released at https://github.com/tmlr-group/Co-rewarding.

cs.LG↗