Flow Map Language Models: One-step Language Modeling via Continuous Denoising

📰 ArXiv cs.AI

Flow Map Language Models achieve faster generation than autoregressive models via continuous denoising

advanced Published 7 Apr 2026

Action Steps

Replace discrete diffusion with continuous flows over one-hot token embeddings
Implement continuous denoising to improve sample quality
Evaluate the performance of the proposed model in the few-step regime
Compare the results with existing autoregressive models

Who Needs to Know This

ML researchers and engineers on a team can benefit from this research as it provides a new approach to language modeling, and software engineers can implement the proposed model in their applications

Key Insight

💡 Continuous flows over one-hot token embeddings can outperform discrete diffusion in language modeling

Key Takeaways

Flow Map Language Models achieve faster generation than autoregressive models via continuous denoising

Full Article

Title: Flow Map Language Models: One-step Language Modeling via Continuous Denoising

Abstract:
arXiv:2602.16813v2 Announce Type: replace-cross Abstract: Language models based on discrete diffusion have attracted widespread interest for their potential to provide faster generation than autoregressive models. Despite their promise, these models typically produce samples whose quality sharply degrades in the few-step regime, preventing a dramatic speedup in practice. Here, we show that language models based on continuous flows over one-hot token embeddings can outperform discrete diffusion i

Read full paper → ← Back to Reads