AdaptFuse: Training-Free Sequential Preference Learning via Externalized Bayesian Inference
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910103351853056 |
|---|---|
| author | Lin, Fangzhou Li, Peiran Xing, Shuo Yang, Siyuan Ge, Qianwen Yamada, Kazunori Zhang, Ziming Zhang, Haichong Tu, Zhengzhong |
| author_facet | Lin, Fangzhou Li, Peiran Xing, Shuo Yang, Siyuan Ge, Qianwen Yamada, Kazunori Zhang, Ziming Zhang, Haichong Tu, Zhengzhong |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_03925 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AdaptFuse: Training-Free Sequential Preference Learning via Externalized Bayesian Inference Lin, Fangzhou Li, Peiran Xing, Shuo Yang, Siyuan Ge, Qianwen Yamada, Kazunori Zhang, Ziming Zhang, Haichong Tu, Zhengzhong Computation and Language Artificial Intelligence 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. |
| title | AdaptFuse: Training-Free Sequential Preference Learning via Externalized Bayesian Inference |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2604.03925 |