ComPO: Community Preferences for Language Model Personalization

• URL: http://arxiv.org/abs/2410.16027v1
• Date: Mon, 21 Oct 2024 14:02:40 GMT
• Title: ComPO: Community Preferences for Language Model Personalization
• Authors: Sachin Kumar, Chan Young Park, Yulia Tsvetkov, Noah A. Smith, Hannaneh Hajishirzi,
• Abstract summary: ComPO is a method to personalize preference optimization in language models. We collect and release ComPRed, a question answering dataset with community-level preferences from Reddit.
• Score: 122.54846260663922
• License:
• Abstract: Conventional algorithms for training language models (LMs) with human feedback rely on preferences that are assumed to account for an "average" user, disregarding subjectivity and finer-grained variations. Recent studies have raised concerns that aggregating such diverse and often contradictory human feedback to finetune models results in generic models that generate outputs not preferred by many user groups, as they tend to average out styles and norms. To address this issue, we draw inspiration from recommendation systems and propose ComPO, a method to personalize preference optimization in LMs by contextualizing the probability distribution of model outputs with the preference provider. Focusing on group-level preferences rather than individuals, we collect and release ComPRed, a question answering dataset with community-level preferences from Reddit. This dataset facilitates studying diversity in preferences without incurring privacy concerns associated with individual feedback. Our experiments reveal that conditioning language models on a community identifier (i.e., subreddit name) during preference tuning substantially enhances model performance. Conversely, replacing this context with random subreddit identifiers significantly diminishes performance, highlighting the effectiveness of our approach in tailoring responses to communities' preferences.

Related papers

• Personalized Preference Fine-tuning of Diffusion Models [75.22218338096316]
  We introduce PPD, a multi-reward optimization objective that aligns diffusion models with personalized preferences. With PPD, a diffusion model learns the individual preferences of a population of users in a few-shot way. Our approach achieves an average win rate of 76% over Stable Cascade, generating images that more accurately reflect specific user preferences.
  arXiv  Detail & Related papers  (2025-01-11T22:38:41Z)
• No Preference Left Behind: Group Distributional Preference Optimization [46.98320272443297]
  Group Distribution Preference Optimization (GDPO) is a novel framework that aligns language models with the distribution of preferences within a group. GDPO calibrates a language model using statistical estimation of the group's belief distribution. GDPO consistently reduces this alignment gap during training.
  arXiv  Detail & Related papers  (2024-12-28T23:30:47Z)
• Few-shot Steerable Alignment: Adapting Rewards and LLM Policies with Neural Processes [50.544186914115045]
  Large language models (LLMs) are increasingly embedded in everyday applications. Ensuring their alignment with the diverse preferences of individual users has become a critical challenge. We present a novel framework for few-shot steerable alignment.
  arXiv  Detail & Related papers  (2024-12-18T16:14:59Z)
• Beyond the Binary: Capturing Diverse Preferences With Reward Regularization [15.518838657050173]
  We argue that this reliance on binary choices does not capture the broader, aggregate preferences of the target user in real-world tasks. We introduce a simple yet effective method that augments existing binary preference datasets with synthetic preference judgments to estimate potential user disagreement.
  arXiv  Detail & Related papers  (2024-12-05T02:35:46Z)
• Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback [87.37721254914476]
  We introduce a routing framework that combines inputs from humans and LMs to achieve better annotation quality. We train a performance prediction model to predict a reward model's performance on an arbitrary combination of human and LM annotations. We show that the selected hybrid mixture achieves better reward model performance compared to using either one exclusively.
  arXiv  Detail & Related papers  (2024-10-24T20:04:15Z)
• PAL: Pluralistic Alignment Framework for Learning from Heterogeneous Preferences [6.398937923320069]
  We propose PAL, a framework to model human preference complementary to existing pretraining strategies. We show that PAL achieves competitive reward model accuracy compared to strong baselines.
  arXiv  Detail & Related papers  (2024-06-12T17:54:54Z)
• Comparing Bad Apples to Good Oranges: Aligning Large Language Models via Joint Preference Optimization [105.3612692153615]
  We propose a new axis based on eliciting preferences jointly over instruction-response pairs. Joint preferences over instruction and response pairs can significantly enhance the alignment of large language models.
  arXiv  Detail & Related papers  (2024-03-31T02:05:40Z)
• Aligning Crowd Feedback via Distributional Preference Reward Modeling [28.754532173765686]
  We propose the Distributional Preference Reward Model (DPRM) to align large language models with diverse human preferences. Our experiments show that DPRM significantly enhances the alignment of LLMs with population preference, yielding more accurate, unbiased, and contextually appropriate responses.
  arXiv  Detail & Related papers  (2024-02-15T07:29:43Z)
• MaxMin-RLHF: Alignment with Diverse Human Preferences [101.57443597426374]
  Reinforcement Learning from Human Feedback (RLHF) aligns language models to human preferences by employing a singular reward model derived from preference data. We learn a mixture of preference distributions via an expectation-maximization algorithm to better represent diverse human preferences. Our algorithm achieves an average improvement of more than 16% in win-rates over conventional RLHF algorithms.
  arXiv  Detail & Related papers  (2024-02-14T03:56:27Z)

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.