Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unlocking the Potential of Large Language Models for Explainable Recommendations

Yucong Luo, Mingyue Cheng, Hao Zhang, Junyu Lu, Qi Liu, Enhong Chen

TL;DR

The paper addresses the challenge of explainable recommendations by investigating how large language models can generate user-friendly explanations without compromising recommendation accuracy. It proposes LLMXRec, a two-stage framework that decouples the recommender from the explanation generator and employs instruction tuning with parameter-efficient methods like LoRA. Through instruction-tuned LLMs and diverse evaluation modalities—automatic discrimination, human judgments, and local attribute prediction—across three public datasets, the approach yields substantial improvements in explanation quality over untuned baselines and competitive performance against strong LLMs. The work demonstrates the feasibility and practicality of using instruction-tuned LLMs for explainable AI in recommender systems, offering a flexible, model-agnostic path toward higher transparency and user trust.

Abstract

Generating user-friendly explanations regarding why an item is recommended has become increasingly common, largely due to advances in language generation technology, which can enhance user trust and facilitate more informed decision-making when using online services. However, existing explainable recommendation systems focus on using small-size language models. It remains uncertain what impact replacing the explanation generator with the recently emerging large language models (LLMs) would have. Can we expect unprecedented results? In this study, we propose LLMXRec, a simple yet effective two-stage explainable recommendation framework aimed at further boosting the explanation quality by employing LLMs. Unlike most existing LLM-based recommendation works, a key characteristic of LLMXRec is its emphasis on the close collaboration between previous recommender models and LLM-based explanation generators. Specifically, by adopting several key fine-tuning techniques, including parameter-efficient instructing tuning and personalized prompt techniques, controllable and fluent explanations can be well generated to achieve the goal of explanation recommendation. Most notably, we provide three different perspectives to evaluate the effectiveness of the explanations. Finally, we conduct extensive experiments over several benchmark recommender models and publicly available datasets. The experimental results not only yield positive results in terms of effectiveness and efficiency but also uncover some previously unknown outcomes. To facilitate further explorations in this area, the full code and detailed original results are open-sourced at https://github.com/GodFire66666/LLM_rec_explanation/.

Unlocking the Potential of Large Language Models for Explainable Recommendations

TL;DR

The paper addresses the challenge of explainable recommendations by investigating how large language models can generate user-friendly explanations without compromising recommendation accuracy. It proposes LLMXRec, a two-stage framework that decouples the recommender from the explanation generator and employs instruction tuning with parameter-efficient methods like LoRA. Through instruction-tuned LLMs and diverse evaluation modalities—automatic discrimination, human judgments, and local attribute prediction—across three public datasets, the approach yields substantial improvements in explanation quality over untuned baselines and competitive performance against strong LLMs. The work demonstrates the feasibility and practicality of using instruction-tuned LLMs for explainable AI in recommender systems, offering a flexible, model-agnostic path toward higher transparency and user trust.

Abstract

Generating user-friendly explanations regarding why an item is recommended has become increasingly common, largely due to advances in language generation technology, which can enhance user trust and facilitate more informed decision-making when using online services. However, existing explainable recommendation systems focus on using small-size language models. It remains uncertain what impact replacing the explanation generator with the recently emerging large language models (LLMs) would have. Can we expect unprecedented results? In this study, we propose LLMXRec, a simple yet effective two-stage explainable recommendation framework aimed at further boosting the explanation quality by employing LLMs. Unlike most existing LLM-based recommendation works, a key characteristic of LLMXRec is its emphasis on the close collaboration between previous recommender models and LLM-based explanation generators. Specifically, by adopting several key fine-tuning techniques, including parameter-efficient instructing tuning and personalized prompt techniques, controllable and fluent explanations can be well generated to achieve the goal of explanation recommendation. Most notably, we provide three different perspectives to evaluate the effectiveness of the explanations. Finally, we conduct extensive experiments over several benchmark recommender models and publicly available datasets. The experimental results not only yield positive results in terms of effectiveness and efficiency but also uncover some previously unknown outcomes. To facilitate further explorations in this area, the full code and detailed original results are open-sourced at https://github.com/GodFire66666/LLM_rec_explanation/.
Paper Structure (31 sections, 5 equations, 5 figures, 4 tables)

This paper contains 31 sections, 5 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Preliminary
  3. Problem Formulation.
  4. Instruction Tuning.
  5. METHODOLOGY
  6. Overview of the Two-stage Framework
  7. Explanable Generator Construction
  8. Foundation Model Selection.
  9. Instruction Template Construction.
  10. Parameter-Efficient Instruction Tuning.
  11. Instruction Tuning Data Construction.
  12. Evaluation of Generated Explanations
  13. Automatic Evaluation with Fine-tuned LLMs as Discriminator.
  14. Manual Evaluation with Scoring Explanation.
  15. Local Evaluation with Attribute Prediction.
  16. ...and 16 more sections

Figures (5)

  • Figure 1: The two-stage framework and evaluation
  • Figure 2: Human evaluations of different LLMs
  • Figure 3: (a) is Win ratio of different historical interaction lengths (compare to length 10). (b) is Win ratio of fine-tuning LLaMA with different data sizes (compared to fine-tuning using 150 data points)
  • Figure 4: Comparison of win ratio with and without added attributes
  • Figure 5: Explanation of LLMs with the above instruction