AI RESEARCH

Score $\times$ Decoder: A Unified View of Unsupervised Inference-Time Scaling for Hallucination Mitigation

arXiv CS.LG

ArXi:2606.00739v1 Announce Type: new Large language models hallucinate even when the answer lies within their parameters. While inference-time scaling can surface this latent knowledge, the most effective methods require supervision: a trained verifier or reward model.