PAL: Pluralistic Alignment Framework for Learning from Heterogeneous Preferences
read the original abstract
Large foundation models pretrained on raw web-scale data are not readily deployable without additional step of extensive alignment to human preferences. Such alignment is typically done by collecting large amounts of pairwise comparisons from humans ("Do you prefer output A or B?") and learning a reward model or a policy with the Bradley-Terry-Luce (BTL) model as a proxy for a human's underlying implicit preferences. These methods generally suffer from assuming a universal preference shared by all humans, which lacks the flexibility of adapting to plurality of opinions and preferences. In this work, we propose PAL, a framework to model human preference complementary to existing pretraining strategies, which incorporates plurality from the ground up. We propose using the ideal point model as a lens to view alignment using preference comparisons. Together with our novel reformulation and using mixture modeling, our framework captures the plurality of population preferences while simultaneously learning a common preference latent space across different preferences, which can few-shot generalize to new, unseen users. Our approach enables us to use the penultimate-layer representation of large foundation models and simple MLP layers to learn reward functions that are on-par with the existing large state-of-the-art reward models, thereby enhancing efficiency of reward modeling significantly. We show that PAL achieves competitive reward model accuracy compared to strong baselines on 1) Language models with Summary dataset ; 2) Image Generative models with Pick-a-Pic dataset ; 3) A new semisynthetic heterogeneous dataset generated using Anthropic Personas. Finally, our experiments also highlight the shortcoming of current preference datasets that are created using rigid rubrics which wash away heterogeneity, and call for more nuanced data collection approaches.
This paper has not been read by Pith yet.
Forward citations
Cited by 7 Pith papers
-
Variance-aware Reward Modeling with Anchor Guidance
Anchor-guided variance-aware reward modeling uses two response-level anchors to resolve non-identifiability in Gaussian models of pluralistic preferences, yielding provable identification, a joint training objective, ...
-
Task-Dependent Evaluation of LLM Output Homogenization: A Taxonomy-Guided Framework
Proposes a task taxonomy for functional diversity in LLM outputs, validates it via user study, introduces targeted sampling to boost diversity only where needed, and presents evidence that the diversity-quality tradeo...
-
Hidden Consensus:Preference-Validity Compression in Human Feedback
Empirical study of Malaysian preference judgments finds that 79% of prompts have multiple majority-supported responses discarded by single-winner aggregation, indicating measurement of argmax rather than plural alignment.
-
Personalized RewardBench: Evaluating Reward Models with Human Aligned Personalization
Personalized RewardBench reveals that state-of-the-art reward models reach only 75.94% accuracy on personalized preferences and shows stronger correlation with downstream BoN and PPO performance than prior benchmarks.
-
RLHF May Not Reflect Genuine Preferences
RLHF preference measurement is a social science validity problem because annotators routinely produce non-attitudes, constructed responses, and artifacts rather than stable values.
-
When to Ask a Question: Understanding Communication Strategies in Generative AI Tools
A tradeoff model shows generative AI can reduce bias against diverse preferences by strategically eliciting information instead of always inferring from majority patterns.
-
POPI: Personalizing LLMs via Optimized Natural Language Preference Inference
POPI distills user preferences into reusable natural-language summaries via a shared inference model and conditions a generator on them, trained jointly with RL to improve personalization quality while cutting context...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.