Related papers: Supervised Fine-Tuning vs. In-Context Learning: An Equilibrium Analysis of LLM Personalization under Congestion

Supervised Fine-Tuning vs. In-Context Learning: An Equilibrium Analysis of LLM Personalization under Congestion

URL: http://arxiv.org/abs/2607.14371v1
Date: Wed, 15 Jul 2026 21:17:42 GMT
Title: Supervised Fine-Tuning vs. In-Context Learning: An Equilibrium Analysis of LLM Personalization under Congestion
Abstract summary: We show that ICL and SFT dominate in different regimes, determined by an interplay between pretraining coverage and data signal-to-noise ratios.<n>We prove that offering both personalization methods never hurts the platform's maximal profits, despite potentially increasing computational load.
Score: 52.722575491987904
License: http://creativecommons.org/licenses/by/4.0/
Abstract: Large Language Models (LLMs) have revolutionized AI services, but a critical tension emerges: while personalization improves model performance, it consumes scarce computational resources that users must share. When should a user invest in expensive Supervised Fine-Tuning (SFT) versus lightweight In-Context Learning (ICL)? How does congestion from other users' personalization choices reshape these incentives? And what strategies should platforms adopt when offering multiple personalization algorithms? We develop a tractable framework for LLM serving that captures the statistical-economic trade-offs users face. Our analysis yields several surprising insights. First, we show that ICL and SFT dominate in different regimes, determined by an interplay between pretraining coverage and data signal-to-noise ratios, but congestion can flip these rankings. Second, equilibrium resource consumption exhibits pronounced non-monotonicity: improving pretraining precision reduces the congestion, while broader pretraining coverage and harder tasks sometimes increase it. Third, we prove that offering both personalization methods never hurts the platform's maximal profits, despite potentially increasing computational load. Experiments with GPT-2 on linear regression tasks validate our theoretical predictions about algorithm performance. Complementing these results, our review of documentation from 21 major AI platforms shows that the share offering both SFT and ICL increased from 9.5% in 2021 to 71.4% in 2025, consistent with our platform-design implications.

Related papers

A Theory of Training Profit-Optimal LLMs [12.802715423118316]
We develop an economic model for characterizing the rational behavior of an LLM training firm.<n>We analyze the profit problem for this model under compute-bound and data-bound regimes.
arXiv Detail & Related papers (2026-05-14T18:57:40Z)
Data Distribution as a Lever for Guiding Optimizers Toward Superior Generalization in LLMs [60.68927774057402]
We show, for the first time, that a lower simplicity bias induces a better generalization.<n>Motivated by this insight, we demonstrate that the training data distribution by upsampling or augmenting examples learned later in training similarly reduces SB and leads to improved generalization.<n>Our strategy improves the performance of multiple language models including Phi2-2.7B, Llama3.2-1B, Gemma3-1B-PT, Qwen3-0.6B-Base-achieving relative accuracy gains up to 18% when fine-tuned with AdamW and Muon.
arXiv Detail & Related papers (2026-01-31T07:40:36Z)
Supervised Fine-Tuning or Contrastive Learning? Towards Better Multimodal LLM Reranking [56.46309219272326]
For large language models (LLMs), classification via supervised fine-tuning (SFT) predicts ''yes'' (resp. ''no'') token for relevant (resp. irrelevant) pairs.<n>This divergence raises a central question: which objective is intrinsically better suited to LLM-based reranking, and what mechanism underlies the difference?<n>We conduct a comprehensive comparison and analysis between CL and SFT for reranking, taking the universal multimodal retrieval (UMR) as the experimental playground.
arXiv Detail & Related papers (2025-10-16T16:02:27Z)
Thinking Augmented Pre-training [88.04395622064708]
Thinking augmented Pre-Training is a universal methodology that augments text with automatically generated thinking trajectories.<n>This paper introduces a simple and scalable approach to improve the data efficiency of large language model (LLM) training by augmenting existing text data with thinking trajectories.
arXiv Detail & Related papers (2025-09-24T14:45:13Z)
Optimal Sparsity of Mixture-of-Experts Language Models for Reasoning Tasks [17.067788440109137]
Mixture-of-Experts (MoE) models are now standard in state-of-the-art systems.<n>We investigate how MoE sparsity influences two distinct capability regimes: memorization skills and reasoning skills.
arXiv Detail & Related papers (2025-08-26T04:31:28Z)
Supervised Fine Tuning on Curated Data is Reinforcement Learning (and can be improved) [3.13388270461847]
We draw on a connection between supervised fine-tuning (SFT) and the theory and practice of finding optimal policies via Reinforcement Learning (RL)<n>We show that a small modification to SFT leads to an importance weighted variant that behaves closer to training with RL as it.<n>We refer to this variant as importance weighted supervised fine-tuning (iw-SFT)
arXiv Detail & Related papers (2025-07-17T07:26:54Z)
Enhancing Training Data Attribution with Representational Optimization [57.61977909113113]
Training data attribution methods aim to measure how training data impacts a model's predictions.<n>We propose AirRep, a representation-based approach that closes this gap by learning task-specific and model-aligned representations explicitly for TDA.<n>AirRep introduces two key innovations: a trainable encoder tuned for attribution quality, and an attention-based pooling mechanism that enables accurate estimation of group-wise influence.
arXiv Detail & Related papers (2025-05-24T05:17:53Z)
Incentivize Contribution and Learn Parameters Too: Federated Learning with Strategic Data Owners [10.49446504694834]
We consider the incentive of an agent for participating and contributing to the process, given that data collection and running a distributed algorithm is costly for the clients.<n>This paper addresses the question of simultaneous parameter learning and incentivizing contribution in a truthful manner, which distinguishes it from the extant literature.<n>Large scale experiments with real (federated) datasets (CIFAR-10, FEMNIST, and Twitter) show that these algorithms converge quite fast in practice, yield good welfare guarantees and better model performance for all agents.
arXiv Detail & Related papers (2025-05-17T14:04:20Z)
CLIP the Bias: How Useful is Balancing Data in Multimodal Learning? [72.19502317793133]
We study the effectiveness of data-balancing for mitigating biases in contrastive language-image pretraining (CLIP) We present a novel algorithm, called Multi-Modal Moment Matching (M4), designed to reduce both representation and association biases.
arXiv Detail & Related papers (2024-03-07T14:43:17Z)
FedCLIP: Fast Generalization and Personalization for CLIP in Federated Learning [18.763298147996238]
Federated learning (FL) has emerged as a new paradigm for privacy-preserving computation in recent years. FL faces two critical challenges that hinder its actual performance: data distribution Heterogeneous and high resource costs. We propose FedCLIP to achieve fast generalization and personalization for CLIP in FL.
arXiv Detail & Related papers (2023-02-27T02:49:06Z)
SCEI: A Smart-Contract Driven Edge Intelligence Framework for IoT Systems [15.796325306292134]
Federated learning (FL) enables collaborative training of a shared model on edge devices while maintaining data privacy. Various personalized approaches have been proposed, but such approaches fail to handle underlying shifts in data distribution. This paper presents a dynamically optimized personal deep learning scheme based on blockchain and federated learning.
arXiv Detail & Related papers (2021-03-12T02:57:05Z)

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