Supervised Fine-Tuning vs. In-Context Learning: An Equilibrium Analysis of LLM Personalization under Congestion
Fengzhuo Zhang, Zhuoran Yang, Dirk Bergemann
cs.LG, econ.TH, stat.ML
Submitted: 2026-07-15
License: http://creativecommons.org/licenses/by/4.0/
The gist: 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
Terminology
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.
Sources
- The Llama 3 Herd of Models
- An Economy of AI Agents
- Unveiling the Statistical Foundations of Chain-of-Thought Prompting Methods
- Learning in Mean Field Games: A Survey
- Minimum-Norm Interpolation Under Covariate Shift
- Artificial Intelligence Index Report 2025
- Artificial Intelligence Index Report 2024
- Isotropic Curvature Model for Understanding Deep Learning Optimization: Is Gradient Orthogonalization Optimal?
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- Muon Outperforms Adam in Tail-End Associative Memory Learning
- Llama-3.1-FoundationAI-SecurityLLM-8B-Instruct Technical Report
- An Explanation of In-context Learning as Implicit Bayesian Inference
- What and How does In-Context Learning Learn? Bayesian Model Averaging, Parameterization, and Generalization
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks