Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding
StepRefine identifies entropy spikes in multimodal diffusion decoding and uses targeted latent refinement to reduce hallucinations without retraining.
ls ./papers | grep "#Latent refinement"
Trending AI research papers from Hugging Face, summarized in plain language: what each paper does, why it matters and where to read more.
1 paper · #Latent refinement
StepRefine identifies entropy spikes in multimodal diffusion decoding and uses targeted latent refinement to reduce hallucinations without retraining.