Training Apr 2026 · #34 most cited · 145 citations

Rethinking On-Policy Distillation of Large Language Models: Phenomenology, Mechanism, and Recipe

Yaxuan Li et al.

arXiv:2604.13016

In short

On-policy distillation, where a student learns from a teacher’s judgement of the student’s own outputs, is widely used but poorly understood. This study finds it works only when teacher and student think in compatible ways and the teacher knows something genuinely new, and offers two fixes for when it fails.

Why it matters

Practical guidance on a post-training technique most labs now depend on.

Read first

The 4 Field Guide ideas this paper leans on.

Starting from scratch? The full route 23 ideas · basics first
  1. Natural Language Processing ✓ understood

    The field of AI that lets computers read, interpret, translate and generate human language, from spam filters and search to chatbots.

  2. Token ✓ understood

    The unit of text a language model reads and writes: a word, part of a word, or a character, drawn from the model's fixed vocabulary.

  3. Tokenization ✓ understood

    Splitting text into tokens, usually subword pieces, and mapping each to an integer ID so a language model can process it.

  4. Language Modeling ✓ understood

    Assigning probabilities to sequences of text, in practice by predicting each next token from the ones before it; the objective behind large language models.

  5. Dataset ✓ understood

    A collection of data examples used for training, validating, or testing machine learning models.

  6. Training ✓ understood

    The process of fitting a model to data by repeatedly measuring how wrong its outputs are and adjusting its parameters to reduce that error.

  7. Machine Learning ✓ understood

    Building systems that learn patterns from data instead of following hand-written rules, getting better at a task as they see more examples.

  8. Unsupervised Learning ✓ understood

    Learning from unlabeled data to discover hidden patterns, structures, or relationships without explicit target outputs.

  9. Self-Supervised Learning ✓ understood

    Training on unlabeled data by making the data supply its own answers: hide or alter part of an input and have the model predict or match it.

  10. Pre-training ✓ understood

    Training a model on a large dataset (often self-supervised) before fine-tuning on specific tasks, enabling transfer learning.

  11. Feature ✓ understood

    A single measurable property of an example, such as a house's floor area or how many links an email contains, used as an input to a model.

  12. Neural Network ✓ understood

    A model built from layers of simple units, each taking a weighted sum of its inputs and applying a nonlinearity; training adjusts the weights so the whole stack maps inputs to useful outputs.

  13. Deep Learning ✓ understood

    A subset of machine learning that uses neural networks with multiple layers (deep neural networks) to learn hierarchical representations of data.

  14. Representation Learning ✓ understood

    Letting a model learn its own internal description of the data, such as embeddings or layered visual features, instead of relying on features designed by hand.

  15. Embedding ✓ understood

    A list of numbers (a vector) that represents a word, sentence, image or other item, learned so that similar items end up close together.

  16. Attention Mechanism ✓ understood

    A technique that lets a neural network weigh every part of its input when producing each output, focusing on the parts most relevant at that step.

  17. Transformer ✓ understood

    A neural network architecture, introduced in 2017, built from stacked self-attention and feed-forward layers; the basis of nearly every modern large language model.

  18. Large Language Model · read first ✓ understood

    A neural network, almost always a transformer, trained on vast amounts of text to predict the next token, which lets it write, answer, summarize and follow instructions.

  19. Activation Function ✓ understood

    A non-linear function applied to neuron outputs that introduces non-linearity, enabling networks to learn complex patterns.

  20. Softmax ✓ understood

    A function that turns a list of scores (logits) into probabilities that are all positive and sum to 1; the standard output of classifiers and language models.

  21. Knowledge Distillation · read first ✓ understood

    Training a smaller 'student' model to mimic a larger 'teacher' model, transferring knowledge while reducing size.

  22. Teacher Model · read first ✓ understood

    The larger, more accurate model in knowledge distillation that guides student training.

  23. Student Model · read first ✓ understood

    The smaller model in knowledge distillation learning to mimic the teacher's behavior.

In the frontier

Rank
#34 of 100
Citations
145
as of Aug 9, 2026
Published
Apr 2026

Topics: Efficiency and serving , Reasoning methods , Data and synthetic generation

Selection: 1kpapers.com by Together AI, most-cited as of Aug 9, 2026

Nearby papers

Summary in our own words; read the paper for the details. ← All papers