HYDRA: Model Factorization Framework for Black-Box LLM Personalization
- URL: http://arxiv.org/abs/2406.02888v2
- Date: Tue, 11 Jun 2024 01:51:57 GMT
- Title: HYDRA: Model Factorization Framework for Black-Box LLM Personalization
- Authors: Yuchen Zhuang, Haotian Sun, Yue Yu, Rushi Qiang, Qifan Wang, Chao Zhang, Bo Dai,
- Abstract summary: Personalization has emerged as a critical research area in modern intelligent systems.
Despite the remarkable few-shot capabilities exhibited by black-box large language models (LLMs), the inherent opacity of their model parameters presents significant challenges in aligning the generated output with individual expectations.
We propose HYDRA, a model factorization framework that captures both user-specific behavior patterns from historical data and shared general knowledge among all users to deliver personalized generation.
- Score: 36.21602686505842
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Personalization has emerged as a critical research area in modern intelligent systems, focusing on mining users' behavioral history and adapting to their preferences for delivering tailored experiences. Despite the remarkable few-shot capabilities exhibited by black-box large language models (LLMs), the inherent opacity of their model parameters presents significant challenges in aligning the generated output with individual expectations. Existing solutions have primarily focused on prompt design to incorporate user-specific profiles and behaviors; however, such approaches often struggle to generalize effectively due to their inability to capture shared knowledge among all users. To address these challenges, we propose HYDRA, a model factorization framework that captures both user-specific behavior patterns from historical data and shared general knowledge among all users to deliver personalized generation. In order to capture user-specific behavior patterns, we first train a reranker to prioritize the most useful information from top-retrieved relevant historical records. By combining the prioritized history with the corresponding query, we train an adapter to align the output with individual user-specific preferences, eliminating the reliance on access to inherent model parameters of black-box LLMs. Both the reranker and the adapter can be decomposed into a base model with multiple user-specific heads, resembling a hydra. The base model maintains shared knowledge across users, while the multiple personal heads capture user-specific preferences. Experimental results demonstrate that HYDRA outperforms existing state-of-the-art prompt-based methods by an average relative improvement of 9.01% across five diverse personalization tasks in the LaMP benchmark. Our implementation is available at https://github.com/night-chen/HYDRA.
Related papers
- PEFT-U: Parameter-Efficient Fine-Tuning for User Personalization [9.594958534074074]
We introduce the PEFT-U Benchmark: a new dataset for building and evaluating NLP models for user personalization.
We explore the challenge of efficiently personalizing LLMs to accommodate user-specific preferences in the context of diverse user-centered tasks.
arXiv Detail & Related papers (2024-07-25T14:36:18Z) - Towards Unified Multi-Modal Personalization: Large Vision-Language Models for Generative Recommendation and Beyond [87.1712108247199]
Our goal is to establish a Unified paradigm for Multi-modal Personalization systems (UniMP)
We develop a generic and personalization generative framework, that can handle a wide range of personalized needs.
Our methodology enhances the capabilities of foundational language models for personalized tasks.
arXiv Detail & Related papers (2024-03-15T20:21:31Z) - Democratizing Large Language Models via Personalized Parameter-Efficient
Fine-tuning [39.0799671550279]
Personalization in large language models (LLMs) is increasingly important.
One PEFT Per User (OPPU) employs personalized parameter-efficient fine-tuning (PEFT) modules to store user-specific behavior patterns and preferences.
OPPU significantly outperforms existing prompt-based methods across seven diverse tasks in the LaMP benchmark.
arXiv Detail & Related papers (2024-02-06T21:03:52Z) - Causal Structure Representation Learning of Confounders in Latent Space
for Recommendation [6.839357057621987]
Inferring user preferences from the historical feedback of users is a valuable problem in recommender systems.
We consider the influence of confounders, disentangle them from user preferences in the latent space, and employ causal graphs to model their interdependencies.
arXiv Detail & Related papers (2023-11-02T08:46:07Z) - Attention Weighted Mixture of Experts with Contrastive Learning for
Personalized Ranking in E-commerce [21.7796124109]
We propose Attention Weighted Mixture of Experts (AW-MoE) with contrastive learning for personalized ranking.
AW-MoE has been successfully deployed in the JD e-commerce search engine.
arXiv Detail & Related papers (2023-06-08T07:59:08Z) - Latent User Intent Modeling for Sequential Recommenders [92.66888409973495]
Sequential recommender models learn to predict the next items a user is likely to interact with based on his/her interaction history on the platform.
Most sequential recommenders however lack a higher-level understanding of user intents, which often drive user behaviors online.
Intent modeling is thus critical for understanding users and optimizing long-term user experience.
arXiv Detail & Related papers (2022-11-17T19:00:24Z) - Incremental user embedding modeling for personalized text classification [12.381095398791352]
Individual user profiles and interaction histories play a significant role in providing customized experiences in real-world applications.
We propose an incremental user embedding modeling approach, in which embeddings of user's recent interaction histories are dynamically integrated into the accumulated history vectors.
We demonstrate the effectiveness of this approach by applying it to a personalized multi-class classification task based on the Reddit dataset.
arXiv Detail & Related papers (2022-02-13T17:33:35Z) - Combining Diverse Feature Priors [90.74601233745047]
We show that models trained with diverse sets of feature priors have less overlapping failure modes.
We also demonstrate that jointly training such models on additional (unlabeled) data allows them to correct each other's mistakes.
arXiv Detail & Related papers (2021-10-15T17:31:10Z) - Unsupervised Model Personalization while Preserving Privacy and
Scalability: An Open Problem [55.21502268698577]
This work investigates the task of unsupervised model personalization, adapted to continually evolving, unlabeled local user images.
We provide a novel Dual User-Adaptation framework (DUA) to explore the problem.
This framework flexibly disentangles user-adaptation into model personalization on the server and local data regularization on the user device.
arXiv Detail & Related papers (2020-03-30T09:35:12Z) - MetaSelector: Meta-Learning for Recommendation with User-Level Adaptive
Model Selection [110.87712780017819]
We propose a meta-learning framework to facilitate user-level adaptive model selection in recommender systems.
We conduct experiments on two public datasets and a real-world production dataset.
arXiv Detail & Related papers (2020-01-22T16:05:01Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.