LLM-Augmented Health Community Notes

Sun, 01 Nov 2026 00:00:00 +0000

Community Notes can provide context for misleading health posts, but useful corrections often arrive late. We study how large language models can support timely, evidence-grounded correction: first by augmenting and automating individual notes, and then by reusing experience from previous cases to improve future notes. The project connects two complementary studies of health misinformation governance on social platforms.

In Beyond the Crowd: LLM-Augmented Community Notes for Governing Health Misinformation (ACL 2026), we analyze 30.8K health-related notes and find a median delay of 17.6 hours before notes receive a helpfulness status. We introduce CrowdNotes+, which combines evidence-grounded note augmentation with utility-guided note automation. Its hierarchical evaluation separates relevance, correctness, and helpfulness, addressing the tendency to mistake fluent writing for factual accuracy. HealthNotes, a benchmark of 1.2K health notes annotated for helpfulness, supports this study. The CrowdNotes+ repository provides the code and data.

In Better with Experience: Self-Evolving LLM Agents for Evidence-Grounded Health Community Notes (Findings of EMNLP 2026), we investigate how note generation can improve across correction episodes. EvoNote distills feedback from completed trajectories into an evolving experience memory that guides claim analysis, evidence acquisition, and note writing. We evaluate it on MM-HealthCN, a 1.2K-instance multimodal benchmark with human-written notes and crowd-derived helpfulness labels. This extends the project’s focus from improving individual corrections to learning reusable correction strategies. The EvoNote repository provides the implementation and MM-HealthCN data.

Better with Experience: Self-Evolving LLM Agents for Evidence-Grounded Health Community Notes

Thu, 01 Oct 2026 00:00:00 +0000

Why Is This Joke Offensive? Interpreting Offensiveness in Chinese Dark Humor Through Knowledge and Rhetoric

Thu, 01 Oct 2026 00:00:00 +0000

Discursive Circuits: How Do Language Models Understand Discourse Relations?

Tue, 04 Nov 2025 00:00:00 +0000

Reasoning and Planning for Large Language Models

Mon, 28 Apr 2025 00:00:00 +0000

Beyond Memorization: The Challenge of Random Memory Access in Language Models

Sun, 11 Aug 2024 00:00:00 +0000

Discursive Socratic Questioning: Evaluating the Faithfulness of Language Models’ Understanding of Discourse Relations