cs.LGNov 19, 2026

PosteriorBench: Evaluating Behavior Posterior Sampling in Language Models

Authors: Zhiyang Xun, Daniel D. Johnson, Sarah Schwettmann, Neil Chowdhury

Organizations: Transluce · The University of Texas at Austin · Thinking Machines Lab

Abstract

Rare language model behaviors can be difficult to study because obtaining them through ordinary sampling can require a very large number of samples. This motivates the behavior posterior sampling problem: given a language model and a behavior of interest, is there an algorithm that efficiently samples from the model's distribution over transcripts that exhibit this behavior? We introduce PosteriorBench, a benchmark for comparing algorithms by their compute cost and by how close their samples are to the true posterior in forward and reverse KL divergence. PosteriorBench contains 21 tasks on Qwen3.5-0.8B with behavior probabilities from 10−510^{-5} to 10−110^{-1}, each with exact posterior samples from a corpus of one billion labeled responses. We find that no method we evaluate is both efficient and faithful: simple methods like prompting are cheap but plateau at a high KL divergence from the posterior, fine-tuning on posterior samples is often cost-prohibitive, and several training methods from prior work do not perform better. Building a posterior sampler that can cheaply approximate the true posterior remains an open problem, and we hope PosteriorBench can measure progress towards this objective. Code and data are available at https://github.com/TransluceAI/PosteriorBench.

Figures & tables

Appendix figures & tables22 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Conformal Language Modeling via Posterior Sampling

    Jun 2, 2026Nicolas Emmenegger, Theo X. Olausson, Armando Solar-Lezama +1Hallucination in Language ModelsConformal Prediction

  2. Representation-Based Exploration for Language Models: From Test-Time to Post-Training

    Oct 13, 2025Jens Tuyls, Dylan J. Foster, Akshay Krishnamurthy +1RL for Language Model ReasoningTest-Time Scaling