Brewing Knowledge in Context: Distillation Perspectives on In-Context Learning

• URL: http://arxiv.org/abs/2506.11516v1
• Date: Fri, 13 Jun 2025 07:17:41 GMT
• Title: Brewing Knowledge in Context: Distillation Perspectives on In-Context Learning
• Authors: Chengye Li, Haiyun Liu, Yuanxi Li,
• Abstract summary: In-context learning (ICL) allows large language models to solve novel tasks without weight updates.<n>Despite its empirical success, the mechanism behind ICL remains poorly understood.<n>We propose a new theoretical perspective that interprets ICL as an implicit form of knowledge distillation.
• Score: 2.6129523822281415
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: In-context learning (ICL) allows large language models (LLMs) to solve novel tasks without weight updates. Despite its empirical success, the mechanism behind ICL remains poorly understood, limiting our ability to interpret, improve, and reliably apply it. In this paper, we propose a new theoretical perspective that interprets ICL as an implicit form of knowledge distillation (KD), where prompt demonstrations guide the model to form a task-specific reference model during inference. Under this view, we derive a Rademacher complexity-based generalization bound and prove that the bias of the distilled weights grows linearly with the Maximum Mean Discrepancy (MMD) between the prompt and target distributions. This theoretical framework explains several empirical phenomena and unifies prior gradient-based and distributional analyses. To the best of our knowledge, this is the first to formalize inference-time attention as a distillation process, which provides theoretical insights for future prompt engineering and automated demonstration selection.

Related papers

• Large Language Models as Computable Approximations to Solomonoff Induction [11.811838796672369]
  We establish the first formal connection between large language models (LLMs) and Algorithmic Information Theory (AIT)<n>We leverage AIT to provide a unified theoretical explanation for in-context learning, few-shot learning, and scaling laws.<n>Our framework bridges the gap between theoretical foundations and practical LLM behaviors, providing both explanatory power and actionable insights for future model development.
  arXiv  Detail & Related papers  (2025-05-21T17:35:08Z)
• Model Steering: Learning with a Reference Model Improves Generalization Bounds and Scaling Laws [52.10468229008941]
  This paper formalizes an emerging learning paradigm that uses a trained model as a reference to guide and enhance the training of a target model through strategic data selection or weighting.<n>We provide theoretical insights into why this approach improves generalization and data efficiency compared to training without a reference model.<n>Building on these insights, we introduce a novel method for Contrastive Language-Image Pretraining with a reference model, termed DRRho-CLIP.
  arXiv  Detail & Related papers  (2025-05-10T16:55:03Z)
• Honey, I Shrunk the Language Model: Impact of Knowledge Distillation Methods on Performance and Explainability [3.224880576815583]
  High computational and storage demands of Large Language Models limit their deployment in resource-constrained environments.<n>Previous research has introduced several distillation methods for both generating training data and for training the student model.<n>Despite their relevance, the effects of state-of-the-art distillation methods on model performance and explainability have not been thoroughly investigated.
  arXiv  Detail & Related papers  (2025-04-22T17:32:48Z)
• Will Pre-Training Ever End? A First Step Toward Next-Generation Foundation MLLMs via Self-Improving Systematic Cognition [86.21199607040147]
  Self-Improving cognition (SIcog) is a self-learning framework for constructing next-generation foundation language models.<n>We introduce Chain-of-Description, a step-by-step visual understanding method, and integrate structured chain-of-thought (CoT) reasoning to support in-depth multimodal reasoning.<n>Extensive experiments demonstrate that SIcog produces next-generation foundation MLLMs with substantially improved multimodal cognition.
  arXiv  Detail & Related papers  (2025-03-16T00:25:13Z)
• Efficient Knowledge Injection in LLMs via Self-Distillation [50.24554628642021]
  This paper proposes utilizing prompt distillation to internalize new factual knowledge from free-form documents.<n>We show that prompt distillation outperforms standard supervised fine-tuning and can even surpass RAG.
  arXiv  Detail & Related papers  (2024-12-19T15:44:01Z)
• Explainability for Large Language Models: A Survey [59.67574757137078]
  Large language models (LLMs) have demonstrated impressive capabilities in natural language processing. This paper introduces a taxonomy of explainability techniques and provides a structured overview of methods for explaining Transformer-based language models.
  arXiv  Detail & Related papers  (2023-09-02T22:14:26Z)
• Understanding Masked Autoencoders via Hierarchical Latent Variable Models [109.35382136147349]
  Masked autoencoder (MAE) has recently achieved prominent success in a variety of vision tasks. Despite the emergence of intriguing empirical observations on MAE, a theoretically principled understanding is still lacking.
  arXiv  Detail & Related papers  (2023-06-08T03:00:10Z)
• Knowledge Distillation Performs Partial Variance Reduction [93.6365393721122]
  Knowledge distillation is a popular approach for enhancing the performance of ''student'' models. The underlying mechanics behind knowledge distillation (KD) are still not fully understood. We show that KD can be interpreted as a novel type of variance reduction mechanism.
  arXiv  Detail & Related papers  (2023-05-27T21:25:55Z)
• A Theory of Emergent In-Context Learning as Implicit Structure Induction [8.17811111226145]
  Scaling large language models leads to an emergent capacity to learn in-context from example demonstrations. We argue that in-context learning relies on recombination of compositional operations found in natural language data. We show how in-context learning is supported by a representation of the input's compositional structure.
  arXiv  Detail & Related papers  (2023-03-14T15:24:05Z)
• Task-Free Continual Learning via Online Discrepancy Distance Learning [11.540150938141034]
  This paper develops a new theoretical analysis framework which provides generalization bounds based on the discrepancy distance between the visited samples and the entire information made available for training the model. Inspired by this theoretical model, we propose a new approach enabled by the dynamic component expansion mechanism for a mixture model, namely the Online Discrepancy Distance Learning (ODDL)
  arXiv  Detail & Related papers  (2022-10-12T20:44:09Z)
• A Free Lunch from the Noise: Provable and Practical Exploration for Representation Learning [55.048010996144036]
  We show that under some noise assumption, we can obtain the linear spectral feature of its corresponding Markov transition operator in closed-form for free. We propose Spectral Dynamics Embedding (SPEDE), which breaks the trade-off and completes optimistic exploration for representation learning by exploiting the structure of the noise.
  arXiv  Detail & Related papers  (2021-11-22T19:24:57Z)
• Understanding Interpretability by generalized distillation in Supervised Classification [3.5473853445215897]
  Recent interpretation strategies focus on human understanding of the underlying decision mechanisms of the complex Machine Learning models. We propose an interpretation-by-distillation formulation that is defined relative to other ML models. We evaluate our proposed framework on the MNIST, Fashion-MNIST and Stanford40 datasets.
  arXiv  Detail & Related papers  (2020-12-05T17:42:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.