# TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation

**Authors:** Keqin Bao, Jizhi Zhang, Yang Zhang, Wenjie Wang, Fuli Feng, Xiangnan He  
**arXiv:** [2305.00447](https://arxiv.org/abs/2305.00447)  
**Date:** 2023-04-30  
**Summary page:** https://www.summarizepaper.com/en/paper/2305.00447v3/

## Key points

- The paper introduces a framework called TALLRec to enhance the recommendation capabilities of Large Language Models (LLMs).
- LLMs have impressive performance in various domains, but their performance in recommendation tasks is suboptimal due to training task disparity and inadequate recommendation data.
- TALLRec builds a Large Recommendation Language Model by tuning LLMs with recommendation data.
- In-context Learning is used to align LLMs with recommendation tasks by formulating the task as prompts for LLMs.
- TALLRec significantly enhances the recommendation capabilities of LLMs in movie and book domains, even with a limited dataset.
- TALLRec is efficient and can be executed on a single RTX 3090 with LLaMA-7B, making it practical for real-world applications.
- The fine-tuned LLM exhibits robust cross-domain generalization, highlighting the effectiveness of TALLRec.
- TALLRec offers an efficient and effective solution for aligning LLMs with recommendation tasks, improving existing systems in diverse domains.

## Summary

The paper titled "TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation" introduces a framework called TALLRec, which aims to enhance the recommendation capabilities of Large Language Models (LLMs). LLMs have shown impressive performance in various domains, leading researchers to explore their potential in recommendation systems. However, the performance of LLMs in recommendation tasks is still suboptimal due to the disparity between their training tasks and recommendation tasks, as well as inadequate recommendation data during pre-training. To address this gap, the authors propose building a Large Recommendation Language Model by tuning LLMs with recommendation data. The TALLRec framework efficiently aligns LLMs with recommendation tasks, leveraging their rich knowledge and strong generalization through In-context Learning. In-context Learning involves formulating the recommendation task as prompts for LLMs. The authors demonstrate that even with a limited dataset of fewer than 100 samples, TALLRec significantly enhances the recommendation capabilities of LLMs in movie and book domains. One notable aspect of TALLRec is its efficiency. The framework can be executed on a single RTX 3090 with LLaMA-7B, making it highly practical for real-world applications. Moreover, the fine-tuned LLM exhibits robust cross-domain generalization, further highlighting the effectiveness of TALLRec. In conclusion, TALLRec offers an efficient and effective solution for aligning LLMs with recommendation tasks. By leveraging recommendation data during tuning, it enhances the performance of LLMs in diverse domains such as movies and books. The proposed framework shows promise for improving existing recommendation systems and can be readily implemented using readily available hardware resources.