Neuron-Aware Data Selection for Annotation-Free LLM Self-Distillation

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Zhuowei Chen, Xiang Lorraine Li
• Venue: arXiv: cs.LG
• Published: 2026-07-02
• Collected: 2026-07-03
• URL: http://arxiv.org/abs/2607.02460v1
• PDF: https://arxiv.org/pdf/2607.02460v1
• DOI: UNKNOWN

Plain-Language Brief

The paper proposes a new method called Neuron On-Policy Self-Distillation (Neuron-OPSD) for annotation-free self-distillation of large language models. This method uses the internal neuron activations to guide data selection and teacher context construction, and it improves in-domain task performance while preserving cross-domain generalization and mitigating calibration collapse over prior annotation-free baselines.

What the Document Says

• The paper discusses a new framework for annotation-free self-distillation of large language models called Neuron On-Policy Self-Distillation (Neuron-OPSD).
• This framework uses internal neuron activations to guide data selection and teacher context construction.
• The model is trained via on-policy distillation from the teacher distribution, requiring no ground-truth labels at any stage.
• Across specialized-domain benchmarks, Neuron-OPSD improves in-domain task performance while preserving cross-domain generalization and mitigating calibration collapse over prior annotation-free baselines.

Key Concepts Mentioned

• Large language models
• Annotation-free self-distillation
• Neuron On-Policy Self-Distillation (Neuron-OPSD)
• Data selection
• Teacher context construction
• On-policy distillation
• In-domain task performance
• Cross-domain generalization
• Calibration collapse

Claims Made by the Document

• Neuron-OPSD improves in-domain task performance.
• Neuron-OPSD preserves cross-domain generalization.
• Neuron-OPSD mitigates calibration collapse.

Evidence Used

• Across specialized-domain benchmarks, Neuron-OPSD improves in-domain task performance while preserving cross-domain generalization and mitigating calibration collapse over prior annotation-free baselines.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Post-training large language models (LLMs) without real-world interaction feedback or human-labeled supervision remains challenging, particularly in specialized domains where expert annotations are costly to obtain. Recent annotation-free self-evolution methods address this by using the model's own outputs as supervision signals, constructing a teacher via additional context and aggregating predictions across multiple rollouts through majority voting to produce pseudo-labels. However, these approaches are not without drawbacks: SFT- and GRPO-based variants suffer out-of-domain performance degradation, while reward-based on-policy RL inflates calibration error. In this paper, we propose Neuron On-Policy Self-Distillation (Neuron-OPSD), a data-centric framework for annotation-free self-distillation that leverages internal neuron activations to guide both training-data selection and teacher context construction. The model is then trained via on-policy distillation from the teacher distribution, requiring no ground-truth labels at any stage. Across specialized-domain benchmarks, Neuron-OPSD improves in-domain task performance while preserving cross-domain generalization and mitigating calibration collapse over prior annotation-free baselines. This framework is particularly relevant to settings where online interaction or external supervision is costly or infeasible, and is conceptually distinct from offline RL approaches that rely on logged, reward-labeled trajectories.