arXiv Machine Learning August 19, 2026 By Jeonghye Kim, Xufang Luo, Minbeom Kim, Sangmook Lee, Dohyung Kim, Jiwon Jeon, Dongsheng Li, Yuqing Yang

Why Does Self-Distillation (Sometimes) Degrade the Reasoning Capability of LLMs?

Read the original on arXiv Machine Learning →

The paper investigates why self‑distillation can sometimes worsen the reasoning abilities of large language models (LLMs). It finds that the process suppresses the model’s epistemic verbalization—its expression of uncertainty—leading to shorter but less accurate responses in mathematical reasoning tasks. Experiments on several LLMs show performance drops of up to 40%, especially on out‑of‑distribution problems where uncertainty expression is beneficial.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

arXiv AI

2d ago

Masked Self-Distillation: Internalizing the Chain-of-Thought in Language Models

arXiv:2607.22629v3 Announce Type: replace Abstract: Large Reasoning Models produce long, explicit chains of intermediate steps before generating a final answer at inference time. These intermediate t...

By Durgesh Kalwar, Vardhan Palod, Jaya Adithya Pavuluri, Subbarao Kambhampati

llms fine-tuning efficiency benchmarks

arXiv AI

3d ago

STEPS: Selective On-Policy Self-Distillation for Reasoning

arXiv:2605.10194v2 Announce Type: replace Abstract: On-policy self-distillation (OPSD) uses a model as its own teacher under privileged context, providing token-level supervision on the model's own r...

By Jiaxuan Wang, Xuan Ouyang, Zhiyu Chen, Yulan Hu, Lan-Zhe Guo

reinforcement-learning efficiency benchmarks

arXiv AI

Jun 10

Adaptive Teacher Exposure for Self-Distillation in LLM Reasoning

arXiv:2605. 11458v3 Announce Type: replace Abstract: On-policy self-distillation has become a strong recipe for LLM reasoning, where a privileged teacher supervises the student's own rollouts while conditioning on the reference solution.

By Zihao Han, Tiangang Zhang, Huaibin Wang, Yilun Sun

llms reinforcement-learning efficiency

arXiv AI

Aug 11

DeltaPrompts: Escaping the Zero-Delta Trap in Multimodal Distillation

arXiv:2605. 15532v3 Announce Type: replace-cross Abstract: Distillation enables compact Vision-Language Models (VLMs) to obtain strong reasoning capabilities, yet the prompts driving this process are typically chosen via simple heuristics or aggregated from off-the-shelf datasets.

By Jaehun Jung, Hyunwoo Kim, Brandon Cui, Ximing Lu, David Acuna, Prithviraj Ammanabrolu, Yejin Choi

llms fine-tuning efficiency multimodal benchmarks

arXiv Machine Learning

Jul 17

On-Policy Delta Distillation

arXiv:2607. 15161v1 Announce Type: new Abstract: On-policy distillation is an alternative post-training method in reinforcement learning that alleviates the constraints imposed by reward models by providing token-level supervision from a teacher model.

By Byeongho Heo, Jaehui Hwang, Sangdoo Yun, Dongyoon Han

llms reinforcement-learning fine-tuning efficiency benchmarks

arXiv Machine Learning

Jun 11

RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation

arXiv:2606. 11709v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) provides dense, token-level supervision for reasoning models by aligning a model's own distribution with the distribution it produces under privileged context, typically a verified solution.

By Leyi Pan, Shuchang Tao, Yunpeng Zhai, Lingzhe Zhang, Zhaoyang Liu, Bolin Ding, Aiwei Liu, Lijie Wen

reinforcement-learning efficiency

One email a morning, machine-written

Related stories

Masked Self-Distillation: Internalizing the Chain-of-Thought in Language Models

STEPS: Selective On-Policy Self-Distillation for Reasoning

Adaptive Teacher Exposure for Self-Distillation in LLM Reasoning

DeltaPrompts: Escaping the Zero-Delta Trap in Multimodal Distillation

On-Policy Delta Distillation

RLCSD: Reinforcement Learning with Contrastive On-Policy Self-Distillation