Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reasoning-Based Personalized Generation for Users with Sparse Data

Bo Ni, Branislav Kveton, Samyadeep Basu, Subhojyoti Mukherjee, Leyao Wang, Franck Dernoncourt, Sungchul Kim, Seunghyun Yoon, Zichao Wang, Ruiyi Zhang, Puneet Mathur, Jihyung Kil, Jiuxiang Gu, Nedim Lipka, Yu Wang, Ryan A. Rossi, Tyler Derr

TL;DR

GraSPer (Graph-based Sparse Personalized Reasoning), a novel framework for enhancing personalized text generation under sparse context, achieves significant performance gain, substantially improving personalization in sparse user context settings.

Abstract

Large Language Model (LLM) personalization holds great promise for tailoring responses by leveraging personal context and history. However, real-world users usually possess sparse interaction histories with limited personal context, such as cold-start users in social platforms and newly registered customers in online E-commerce platforms, compromising the LLM-based personalized generation. To address this challenge, we introduce GraSPer (Graph-based Sparse Personalized Reasoning), a novel framework for enhancing personalized text generation under sparse context. GraSPer first augments user context by predicting items that the user would likely interact with in the future. With reasoning alignment, it then generates texts for these interactions to enrich the augmented context. In the end, it generates personalized outputs conditioned on both the real and synthetic histories, ensuring alignment with user style and preferences. Extensive experiments on three benchmark personalized generation datasets show that GraSPer achieves significant performance gain, substantially improving personalization in sparse user context settings.

Reasoning-Based Personalized Generation for Users with Sparse Data

TL;DR

GraSPer (Graph-based Sparse Personalized Reasoning), a novel framework for enhancing personalized text generation under sparse context, achieves significant performance gain, substantially improving personalization in sparse user context settings.

Abstract

Large Language Model (LLM) personalization holds great promise for tailoring responses by leveraging personal context and history. However, real-world users usually possess sparse interaction histories with limited personal context, such as cold-start users in social platforms and newly registered customers in online E-commerce platforms, compromising the LLM-based personalized generation. To address this challenge, we introduce GraSPer (Graph-based Sparse Personalized Reasoning), a novel framework for enhancing personalized text generation under sparse context. GraSPer first augments user context by predicting items that the user would likely interact with in the future. With reasoning alignment, it then generates texts for these interactions to enrich the augmented context. In the end, it generates personalized outputs conditioned on both the real and synthetic histories, ensuring alignment with user style and preferences. Extensive experiments on three benchmark personalized generation datasets show that GraSPer achieves significant performance gain, substantially improving personalization in sparse user context settings.
Paper Structure (47 sections, 1 theorem, 18 equations, 3 figures, 13 tables, 2 algorithms)

This paper contains 47 sections, 1 theorem, 18 equations, 3 figures, 13 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Problem Definition
  3. Personalized Text Generation with Graph.
  4. Method
  5. Step 1: Personal Context Expansion
  6. Training Graph-based Link Predictor.
  7. Inference and Profile Augmentation.
  8. Step 2: Synthetic Text Generation with Reasoning Alignment
  9. Synthetic Text Generation Setup.
  10. Reasoning Path Generation.
  11. Reasoning Alignment.
  12. Step 3: Personalized Text Generation
  13. Experiments
  14. Datasets and Metrics
  15. Baselines
  16. ...and 32 more sections

Key Result

Proposition 1

Let $\theta_u \in \mathbb{R}^d$ denote the user's latent style vector. We observe $n$ real samples $x_j=\theta_u+\varepsilon_j$ with $\mathbb{E}[\varepsilon_j]=0$, $\mathrm{Var}(\varepsilon_j)=\sigma^2 I$, and $k$ synthetic samples $\tilde{x}_\ell=\theta_u+\Delta+\tilde{\varepsilon}_\ell$ with $\mat Then the (per-coordinate) mean squared error is In the equal-noise case $\tilde{\sigma}^2=\sigma^2

Figures (3)

  • Figure 1: Results comparing our approach across two fundamental tasks and across datasets. Legends use descriptive labels: Ours (Graph + Reasoning), PGraph (Graph Only), LaMP (No Graph, No Reasoning), and REST-PG (Reasoning Only). GraSPeR achieves over 10% gains, on average, across datasets.
  • Figure 2: Overview of the proposed GraSPeR. In Step 1, the personal context is enriched by leveraging the underlying graph structure to predict potential interactions. In Step 2, we generate synthetic reviews for the predicted interactions with aligned reasoning. In Step 3, the enhanced personal context that contains both the observed and simulated interactions enables the generation of more accurate and personalized text.
  • Figure 3: Case study with three examples. The matching green, blue, and yellow boxes show matching semantics or expression. The red box shows misalignment against the ground truth.

Theorems & Definitions (4)

  • Proposition 1: Bias--Variance trade-off
  • proof : Sketch
  • Remark 1: Effect of Reasoning Alignment
  • Remark 2: Sparse Users Benefit More