Training In-Context and In-Weights Mixtures Via Contrastive Context Sampling

• URL: http://arxiv.org/abs/2604.01601v1
• Date: Thu, 02 Apr 2026 04:21:42 GMT
• Title: Training In-Context and In-Weights Mixtures Via Contrastive Context Sampling
• Abstract summary: We investigate training strategies that co-develop in-context learning (ICL) and in-weights learning (IWL)<n>We show that the similarity structure between target inputs and context examples also plays an important role.
• Score: 17.901170420482618
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We investigate training strategies that co-develop in-context learning (ICL) and in-weights learning (IWL), and the ability to switch between them based on context relevance. Although current LLMs exhibit both modes, standard task-specific fine-tuning often erodes ICL, motivating IC-Train - fine-tuning with in-context examples. Prior work has shown that emergence of ICL after IC-Train depends on factors such as task diversity and training duration. In this paper we show that the similarity structure between target inputs and context examples also plays an important role. Random context leads to loss of ICL and IWL dominance, while only similar examples in context causes ICL to degenerate to copying labels without regard to relevance. To address this, we propose a simple Contrastive-Context which enforces two types of contrasts: (1) mix of similar and random examples within a context to evolve a correct form of ICL, and (2) varying grades of similarity across contexts to evolve ICL-IWL mixtures. We present insights on the importance of such contrast with theoretical analysis of a minimal model. We validate with extensive empirical evaluation on four LLMs and several tasks. Diagnostic probes confirm that contrasted contexts yield stable ICL-IWL mixtures, avoiding collapse into pure ICL, IWL, or copying.

Related papers

• TwinICL: Diagnosing Multimodal In-Context Learning through Paired Counterfactuals [49.800443965874756]
  In-context learning (ICL) enables models to infer tasks from demonstrations.<n>Existing benchmarks generally lack matched text and image versions needed to compare ICL performance across modalities.<n>We introduce TwinICL, a procedurally generated benchmark providing such pairs for controlled comparison.
  arXiv  Detail & Related papers  (2026-09-14T04:42:42Z)
• In-Context Learning Is Provably Bayesian Inference: A Generalization Theory for Meta-Learning [51.56484100374058]
  We introduce a principled risk decomposition that separates the total ICL risk into two components: Bayes Gap and Posterior Variance.<n>For a uniform-attention Transformer, we derive a non-asymptotic upper bound on this gap, which explicitly clarifies the dependence on the number of pretraining prompts.<n>The Posterior Variance is a model-independent risk representing the intrinsic task uncertainty.
  arXiv  Detail & Related papers  (2025-10-13T03:42:31Z)
• Strategy Coopetition Explains the Emergence and Transience of In-Context Learning [24.63934469340368]
  In-context learning (ICL) is a powerful ability that emerges in transformer models, enabling them to learn from context without weight updates.<n>Recent work has established emergent ICL as a transient phenomenon that can sometimes disappear after long training times.<n>We propose a minimal mathematical model that reproduces these key dynamics and interactions.
  arXiv  Detail & Related papers  (2025-03-07T17:54:05Z)
• Multimodal Contrastive In-Context Learning [0.9120312014267044]
  This paper introduces a novel multimodal contrastive in-context learning framework to enhance our understanding of gradient-free in-context learning (ICL) in Large Language Models (LLMs) First, we present a contrastive learning-based interpretation of ICL in real-world settings, marking the distance of the key-value representation as the differentiator in ICL. Second, we develop an analytical framework to address biases in multimodal input formatting for real-world datasets. Third, we propose an on-the-fly approach for ICL that demonstrates effectiveness in detecting hateful memes.
  arXiv  Detail & Related papers  (2024-08-23T10:10:01Z)
• ICLEval: Evaluating In-Context Learning Ability of Large Language Models [68.7494310749199]
  In-Context Learning (ICL) is a critical capability of Large Language Models (LLMs) as it empowers them to comprehend and reason across interconnected inputs.<n>Existing evaluation frameworks primarily focus on language abilities and knowledge, often overlooking the assessment of ICL ability.<n>We introduce the ICLEval benchmark to evaluate the ICL abilities of LLMs, which encompasses two key sub-abilities: exact copying and rule learning.
  arXiv  Detail & Related papers  (2024-06-21T08:06:10Z)
• Investigating the Pre-Training Dynamics of In-Context Learning: Task Recognition vs. Task Learning [99.05401042153214]
  In-context learning (ICL) is potentially attributed to two major abilities: task recognition (TR) and task learning (TL) We take the first step by examining the pre-training dynamics of the emergence of ICL. We propose a simple yet effective method to better integrate these two abilities for ICL at inference time.
  arXiv  Detail & Related papers  (2024-06-20T06:37:47Z)
• Implicit In-context Learning [37.0562059811099]
  We introduce Implicit In-context Learning (I2CL), an innovative paradigm that reduces the inference cost of ICL to that of zero-shot learning with minimal information loss.<n>I2CL achieves few-shot level performance at zero-shot inference cost, and it exhibits robustness against variations in demonstration examples.
  arXiv  Detail & Related papers  (2024-05-23T14:57:52Z)
• Investigating the Learning Behaviour of In-context Learning: A Comparison with Supervised Learning [67.25698169440818]
  Large language models (LLMs) have shown remarkable capacity for in-context learning (ICL) We train the same LLMs with the same demonstration examples via ICL and supervised learning (SL), respectively, and investigate their performance under label perturbations. First, we find that gold labels have significant impacts on the downstream in-context performance, especially for large language models. Second, when comparing with SL, we show empirically that ICL is less sensitive to label perturbations than SL, and ICL gradually attains comparable performance to SL as the model size increases.
  arXiv  Detail & Related papers  (2023-07-28T09:03:19Z)
• In-Context Learning Learns Label Relationships but Is Not Conventional Learning [60.891931501449726]
  There is currently no consensus about how in-context learning (ICL) ability of Large Language Models works. We provide novel insights into how ICL leverages label information, revealing both capabilities and limitations. Our experiments show that ICL predictions almost always depend on in-context labels and that ICL can learn truly novel tasks in-context.
  arXiv  Detail & Related papers  (2023-07-23T16:54:41Z)

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.