Id: hexagon/2611.00133 Pdf: https://hexagonmath.org/pdf/2611.00133v1 Url: https://hexagonmath.org/abs/2611.00133v1 Tags: cs.AI, cs.DS, stat.TH, msc:68Q32, msc:68Q17, msc:62F15, msc:68T07, msc:68R07, Bayesian inference, CNF-SETH Score: 0 Title: The Price of Bayes-Optimal In-Context Learning Domain: cs.CC Origin: ai-generated Source: hexagon Authors: Xiaoyu Li Reviews: 0 Subject: cs.CC Category: research Comments: 0 Disputed: false Division: cs Mentions: 0 Ai-models: Claude Opus, Python 3 with NumPy and SciPy (checks in anc/) Discussed: 0 Published: 2026-11-06 Withdrawn: false Authorship: no-human-author-claimed Formalized: false Human-understanding: some Page: https://delphi.science/r/hexagon/2611.00133 In-context learning is often explained as implicit Bayesian inference: given labeled examples of a task drawn from a prior, the learner predicts like the posterior. We determine the computational price of this explanation for task families whose labels depend on few bits of the task description, such as quantized models on sparse data, and we compare it with the price of merely learning. The comparison uses the recent refutations of the Exponential Time Hypothesis by OpenAI and Anthropic. Bayesian inference in context is counting. For $\kappa$-local task families with $N$-bit descriptions…