AdaptFuse: Training-Free Sequential Preference Learning via Externalized Bayesian Inference

TL;DR

AdaptFuse introduces a training-free, externalized Bayesian inference framework for LLMs that improves multi-round evidence accumulation without fine-tuning or sensitive data storage.

Contribution

It presents a novel externalized Bayesian inference method combining symbolic probabilistic modules with frozen LLMs, enabling effective belief updating at inference time.

Findings

AdaptFuse outperforms prompting baselines and fine-tuned models across multiple domains.

Accuracy improves monotonically over interaction rounds.

The method works across various LLM sizes and tasks.

Abstract

Large language models struggle to accumulate evidence across multiple rounds of user interaction, failing to update their beliefs in a manner consistent with Bayesian inference. Existing solutions require fine-tuning on sensitive user interaction data, limiting their applicability in privacy-conscious settings. We propose AdaptFuse, a training-free framework that externalizes probabilistic computation entirely from the LLM: a symbolic module maintains a Bayesian posterior over a discrete hypothesis set, while a frozen LLM contributes semantic reasoning via multi-sample Dirichlet aggregation. The two signals are combined through entropy-adaptive fusion, which automatically weights each source by its predictive confidence, shifting reliance from the LLM to the symbolic posterior as evidence accumulates. We evaluate across three domains: flight recommendation, hotel recommendation, and web shopping; on Gemma 2 9B, Llama 3 8B, and Qwen 2.5 7B. AdaptFuse consistently outperforms both prompting baselines and fine-tuned Bayesian Teaching models on all tasks, with accuracy improving monotonically over interaction rounds. These results demonstrate that principled inference-time algorithms can substitute for fine-tuning in personalized recommendation, without storing or training on sensitive user data. All the code and materials will be open-sourced.