Data Party THU

Oct 23, 2025 · Artificial Intelligence

How Veri‑R1 Enables Reliable Online Fact‑Checking for Large Language Models

This article introduces Veri‑R1, a novel training framework that equips large language models with online claim verification capabilities by combining reinforcement learning, fine‑grained reward design, and high‑quality data filtering, and demonstrates its superior performance across multiple fact‑checking benchmarks.