{"slug": "recipes-for-steering-and-scaling-llms-via-sampling", "title": "Recipes for Steering and Scaling LLMs via Sampling", "summary": "Researchers introduced a framework for steering and scaling autoregressive large language models via sampling, with two algorithms based on Sequential Monte Carlo and Replica Exchange that target powering, product, or tilting of the base distribution. Experiments show the methods scale more favorably than Best-of-N and standard MCMC baselines, improving generation quality without external supervision or reward models.", "body_md": "arXiv:2608.26120v1 Announce Type: new\nAbstract: Large Language Models (LLMs) are probabilistic models, typically defined by an autoregressive factorization. While recent work has begun to study richer target distributions beyond the base model, the sampling strategies remain highly inefficient. In this paper, we present a flexible and theoretically grounded framework for steering and scaling autoregressive LLMs with sampling. Within this framework, we describe two algorithms -- one based on Sequential Monte Carlo (SMC) and one based on Replica Exchange (RE) -- that steer generation toward powering, product or tilting of the base model distribution. We illustrate this framework through scaling the generation quality of LLMs without external supervision or reward models. Experimental results demonstrate our methods scale more favorably than Best-of-N and standard MCMC baselines. Overall, this paper offers a systematic recipe for probabilistic inference with LLMs via sampling.", "url": "https://wpnews.pro/news/recipes-for-steering-and-scaling-llms-via-sampling", "canonical_source": "https://arxiv.org/abs/2608.26120", "published_at": "2026-08-28 04:00:00+00:00", "updated_at": "2026-08-28 04:20:03.060884+00:00", "lang": "en", "topics": ["large-language-models", "ai-research"], "entities": [], "alternates": {"html": "https://wpnews.pro/news/recipes-for-steering-and-scaling-llms-via-sampling", "markdown": "https://wpnews.pro/news/recipes-for-steering-and-scaling-llms-via-sampling.md", "text": "https://wpnews.pro/news/recipes-for-steering-and-scaling-llms-via-sampling.txt", "jsonld": "https://wpnews.pro/news/recipes-for-steering-and-scaling-llms-via-sampling.jsonld"}}