# LLM-Derived Priors for Thompson Sampling in Cold-Start Comment Recommendation

arXiv: https://arxiv.org/abs/2608.03382 · HTML: https://mlsanity.com/paper/2608.03382/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 19 references that are in the map.
- First posted on arXiv 4 Aug 2026.

## Builds on (most cited first, 9 of 9)

- [A Tutorial on Thompson Sampling](https://mlsanity.com/paper/1707.02038/) (2017)
- [A Survey on Large Language Models for Recommendation](https://mlsanity.com/paper/2305.19860/) (2023)
- [Large Language Models are Zero-Shot Rankers for Recommender Systems](https://mlsanity.com/paper/2305.08845/) (2023)
- [Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting](https://mlsanity.com/paper/2306.17563/) (2023)
- [Joint Deep Modeling of Users and Items Using Reviews for Recommendation](https://mlsanity.com/paper/1701.04783/) (2017)
- [The Unfairness of Popularity Bias in Recommendation](https://mlsanity.com/paper/1907.13286/) (2019)
- [A Survey on Popularity Bias in Recommender Systems](https://mlsanity.com/paper/2308.01118/) (2023)
- [How Useful are Reviews for Recommendation? A Critical Review and Potential Improvements](https://mlsanity.com/paper/2005.12210/) (2020)
- [Jump Starting Bandits with LLM-Generated Prior Knowledge](https://mlsanity.com/paper/2406.19317/) (2024)

## Abstract (arXiv)

Multi-armed bandit algorithms, especially Thompson sampling, are widely used in online recommendation. Despite their ability to adapt from online feedback, these methods often suffer from cold-start limitations when newly introduced arms have little or no interaction history. In our setting, the candidate arms are user-generated textual comments, whose semantic content can reveal a title's appeal before sufficient interaction feedback is available. We therefore use large language models (LLMs) to extract semantic signals from comment text and convert them into informative Bayesian priors that warm-start Thompson sampling under sparse early-stage feedback. To account for aggregate segment-level differences in response patterns, we maintain and update posteriors separately for each gender-age segment. In a real-world online A/B/C test, we compare a uniform prior with two LLM-based designs: a Gender Prior for demographic-affinity cues and a Content Prior for title-specific identity cues. The results show that LLM-based priors are most beneficial in sparse-feedback regimes -- with the largest gains emerging once a small amount of interaction evidence has accumulated -- and that prior design leads to distinct funnel-level effects. We further analyze prior-reward alignment and demographic heterogeneity, finding that click-oriented alignment is strongest for the Gender Prior and that treatment effects vary substantially across demographic segments. These findings suggest that LLM-derived priors can serve as a practical warm-start mechanism for text-rich bandit recommendation, while also revealing deployment trade-offs.