AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

In-Context Learning Is Provably Bayesian Inference: A Generalization Theory for Meta-Learning

Tomoya Wakayama, Taiji Suzuki

ICML 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Bayesian view of meta-learning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fit Bayes gap jointly in pretraining prompts N and context length p rather than N only (ablation)power-law surrogate of Bayes gap · Pooled Bayes-gap fit (two-family synthetic)R^2: 0.030 -> 0.912Improved
fit Bayes gap jointly in N and p rather than p only (ablation)power-law surrogate of Bayes gap · Pooled Bayes-gap fit (two-family synthetic)R^2: 0.016 -> 0.912Improved
increase number of pretraining prompts NGPT-2 transformer · Two-family synthetic regressionBayes gapImproved
increase pretraining prompt length pGPT-2 transformer · Two-family synthetic regressionBayes gapImproved
increase observation noise from 0.1 to 0.5Bayes posterior · Three-family mixture, k=3wrong-family posterior mass: 0.067 -> 0.093Hurt
increase observation noise from 0.1 to 0.5Bayes posterior · Three-family mixture, k=5wrong-family posterior mass: 0.0010 -> 0.0054Hurt
increase context length from k=1 to k=5Transformer · Three-family mixture, noise 0.1MSE: 19.15 -> 0.156Improved
increase context length from k=1 to k=5Transformer · Three-family mixture, noise 0.5MSE: 23.94 -> 0.338Improved
give true task family to the Bayes predictor (oracle vs mixture)Bayes predictor · Three-family mixture, noise 0.1, k=1MSE: 18.027 -> 4.063Improved
give true task family to the Bayes predictor (oracle vs mixture)Bayes predictor · Three-family mixture, noise 0.5, k=5MSE: 0.317 -> 0.314No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy