Skip to content

Mitigating the Length-Scaling Tax with Online Distillation

Sep 2026 · 0 citations · 29 references
Computer Science

TL;DR

LSD is proposed, which routes solved prompts to on-policy distillation and retains the original RL objective for unsolved prompts and achieves comparable or better performance than RL across multiple variants, while substantially curbing response-length growth on easy queries.

Abstract

Length scaling during reinforcement-learning (RL) post-training is often viewed as a sign of improved reasoning ability, especially on difficult problems, but may also make responses to already-solved problems unnecessarily verbose. We quantify this side effect as the length-scaling tax (LST): excess response length on already-solved queries without a commensurate accuracy gain. To mitigate LST, we propose Length Self-Distillation (LSD), which routes solved prompts to on-policy distillation and retains the original RL objective for unsolved prompts. LSD uses an exponential moving average of the online policy as its teacher, requiring no external model. We find that LSD achieves comparable or better performance than RL across multiple variants, while substantially curbing response-length growth on easy queries. LSD reduces LST from 19.0% to -3.7% on single-turn reasoning and from 31.4% to 13.7% on multi-turn agentic tasks, demonstrating that LSD effectively preserves concise response patterns on easy queries while supporting efficient exploration on difficult queries during RL post-training.

View source

Similar papers

#artificial intelligence Preprint Oct 2026

Does self-distillation improve reasoning?

This paper revisits two representative methods, Self-Distillation Policy Optimization (SDPO) and On-Policy Self-Distillation (OPSD), and finds that their reported gains do not consistently generalize to unseen problems and are largely explained by overfitting or transferring existing reasoning behavior, rather than eff...

J. Adhinarta, Raina Wu, Ashutosh Sharma et al. · 0 citations
#machine learning Preprint Oct 2026

Why On-Policy Distillation Sometimes Fails: Vanishing Learning Signals

A local recovery guarantee for teachers sufficiently close to the initial student in a shared parameterization under regularity conditions is proved, offering a conditional explanation for the success of self-RL teachers in the authors' experiments.

Lei Zhao, Qi-Chao Zhao, Bo-Wen Zuo et al. · 0 citations
#machine learning Preprint Oct 2026

Outcome-Guided On-Policy Self-Distillation

On-policy self-distillation (OPSD) provides denser token-level supervision and better computational efficiency than Reinforcement Learning with Verifiable Rewards (RLVR). However, this denser supervision may introduce substantial noise and training instability. Existing improvements often rely on high-variance per-toke...

Zhe-Xu Wang, Mao-Lin Luo, Yan-Kun Hong et al. · 0 citations
#artificial intelligence Preprint Oct 2026

Sharpening Tax in Post-Training

An emerging hypothesis about reinforcement learning (RL) post-training of large language models (LLMs) is that it merely sharpens existing behaviors of a base model, improving single-shot accuracy at the cost of solution coverage. Although this trade-off has been observed in math and coding tasks, it need not extend to...

Changdae Oh, Qi Zeng, Qi Qi et al. · 1 citation
#machine learning Preprint Sep 2026

Scaling Properties of Same-Family On-Policy Distillation

It is found that early OPD training dynamics uniformly exhibit a regular *useful-transfer* regime, in which held-out accuracy (the *gold score*, $G$) rises approximately linearly in $d=\sqrt{\mathrm{KL}(\pi_\theta \Vert \pi_{\mathrm{ref}})}$, the square root of token-level reverse KL divergence from the student initial...

Yun-Tai Bao, Qin-Feng Li, Guo-qing Jiang et al. · 0 citations

Related blog posts

MIT News · Artificial Intelligence Sep 29, 2026

Who we become when we talk to machines

Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.