Scaling Human Judgment in Community Notes with LLMs

2025-06-30

AI Summary

This work addresses the challenge of constructing an open-community note ecosystem in the LLM era. We propose a human-in-the-loop, model-assisted annotation framework wherein humans retain ultimate authority over note visibility, while LLMs efficiently generate draft annotations. Our key contribution is the first introduction of “Reinforcement Learning from Community Feedback” (RLCF), a novel mechanism that directly translates heterogeneous, multi-source real-user feedback—such as adoption, editing, or rejection of generated notes—into reinforcement signals to close the optimization loop. By jointly training the model via human collaborative annotation and RLCF, we significantly improve annotation accuracy, fairness, and practical utility. Empirical evaluation demonstrates that our approach accelerates annotation delivery while preserving community trust and governance legitimacy, thereby enabling bidirectional human-AI empowerment and sustainable ecosystem evolution.

Technology Category

• Humans and AI: Human-in-the-loop Machine Learning
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

This paper argues for a new paradigm for Community Notes in the LLM era: an open ecosystem where both humans and LLMs can write notes, and the decision of which notes are helpful enough to show remains in the hands of humans. This approach can accelerate the delivery of notes, while maintaining trust and legitimacy through Community Notes' foundational principle: A community of diverse human raters collectively serve as the ultimate evaluator and arbiter of what is helpful. Further, the feedback from this diverse community can be used to improve LLMs' ability to produce accurate, unbiased, broadly helpful notes--what we term Reinforcement Learning from Community Feedback (RLCF). This becomes a two-way street: LLMs serve as an asset to humans--helping deliver context quickly and with minimal effort--while human feedback, in turn, enhances the performance of LLMs. This paper describes how such a system can work, its benefits, key new risks and challenges it introduces, and a research agenda to solve those challenges and realize the potential of this approach.

Problem

• Balancing human and LLM contributions in Community Notes
• Ensuring note quality via diverse human evaluators
• Improving LLMs through community feedback (RLCF)

Innovation

• Human-LLM collaboration in note creation
• Reinforcement Learning from Community Feedback
• Diverse human raters as ultimate evaluators