Can I read Reasoning Models Hallucinate More: Factuality-Aware Reinforcement Learning For Large Reasoning Models on EtoBox?

Reasoning Models Hallucinate More: Factuality-Aware Reinforcement Learning For Large Reasoning Models by iamtensorai is a document available to read on EtoBox.

What is Reasoning Models Hallucinate More: Factuality-Aware Reinforcement Learning For Large Reasoning Models about?

This document discusses the challenges of hallucinations in large language models (LLMs) that arise from reinforcement learning (RL) fine-tuning, which can lead to factually incorrect outputs during reasoning tasks. The authors propose a new algorithm called Factuality-aware Step-wise Policy Optimization (FSPO) that incorporates factuality verification at each reasoning step to mitigate hallucinations and improve reasoning accuracy. Experimental results demonstrate that FSPO significantly reduces hallucinat

Author

iamtensorai

Language

EN