bayesian

Stanford researchers develop a Bayesian variance-decomposition framework using Generalizability theory to diagnose reliability in agent benchmarks, showing that adding more tasks cannot fix uncertainty caused by scaffold coverage and that pooling diverse benchmarks is a cheaper path to reliable rankings.

This paper proposes an Infinite-Parameter LLM architecture that uses a hypernetwork to generate weights from live data via Bayesian updates, enabling continuous adaptation and outperforming in-context learning.

This paper proposes a unified definition of uncertainty as pointwise posterior risk and introduces a theory-backed benchmark using semi-synthetic datasets to directly compute oracle epistemic and aleatoric uncertainty, enabling fine-grained evaluation beyond proxy tasks.

AgentOPSD proposes a critic-free recursive method for turn-level credit assignment in agentic RL, reweighting outcomes via Bayesian belief updates in log-odds space. It achieves 89.1% success on ALFWorld with Qwen2.5-7B, outperforming GRPO and self-distillation baselines.

MetaKoopman proposes a Bayesian meta-learning framework for modeling nonlinear dynamics using linear latent representations via Koopman operators, enabling closed-form updates and uncertainty quantification. It is validated on autonomous truck and trailer systems under adverse winter conditions, outperforming prior methods in prediction accuracy and robustness.

This paper presents a large-scale study comparing Bayesian complete-pooling models to frequentist baselines for cross-subject motor imagery EEG classification, finding that Bayesian methods improve reliability modestly but at a higher computational cost, with limited practical benefit.

A tutorial overview of simulation-based inference (SBI) with machine learning, covering Bayesian and frequentist frameworks, methods like neural posterior estimation, and validation techniques.

A classic 1982 paper by A. P. Dawid on the concept of well-calibrated Bayesian probability forecasts, foundational in statistics and machine learning.

This paper presents a Bayesian model of intercomprehension—understanding a related language without training—using a noisy-channel approach. It compares the model's predictions to human behavior across three language pairs, showing better alignment than larger zero-shot models.

The author uses a photo of scuff marks on a subway station wall to estimate the height distribution of commuters, applying image processing and a heuristic body-to-scuff ratio, and discusses potential improvements using Bayesian methods.

This paper introduces analytically tractable Bayesian information restricted diffusion (BIRD) models to study the memorization-generalization phase transition in diffusion models, finding that generation proceeds near the edge of memorization and that information restriction helps circumvent the curse of dimensionality.

Proves a tight approximation ratio for the greedy algorithm in myopic Bayesian active learning for linear regression, identifying the maximum initial leverage score as a key quantity.

This paper introduces Manana, a non-parametric prompt-learning framework that teaches LLMs to recommend anti-seizure medications and defer uncertain cases in underrepresented epilepsy care settings, improving accuracy on Ugandan cohorts and enabling selective prediction with high precision.

This paper presents a causal machine learning approach combining double/debiased machine learning with a hierarchical Bayesian framework to estimate the incremental impact of additional supply on marketplace outcomes, using Airbnb as a case study.

Proposes a new RANSAC scoring function that marginalizes the inlier scale analytically, removing the need for user-supplied parameters. The method achieves state-of-the-art accuracy on a benchmark of nearly 70,000 image pairs.

A comprehensive survey on uncertainty quantification in symbolic regression, reviewing frequentist, Bayesian, and model selection approaches to address the lack of reliability support in real-world decision processes.

PBSD proposes a Bayesian self-distillation method that converts sparse final rewards into calibrated turn-level credit signals for long-horizon agentic tasks, improving policy learning and generalization.

This paper presents the first implementation of an infra-Bayesian reinforcement learning agent, demonstrating that it outperforms classical RL in worst-case regret and handles Newcomb's problem optimally, offering a step toward robustness under model misspecification.

This paper proposes a model-agnostic probabilistic token attribution measure for LLMs using Bayes' rule to invert next-token log probabilities, capturing the model's internal representation of token sequences and improving interpretability through entropy analysis.

The paper introduces the Bayesian Filtering Transformer (BFT), which incorporates uncertainty into Transformers via precision-weighted attention and Kalman update residuals, improving performance on sequential recommendation and noisy LLM fine-tuning.