NeurIPS 2026 Workshop to Explore Discrete Diffusion and Non-Causal Generation

ArashVahdat · x · 2026-08-13

The BeNTo (Beyond Next Token Prediction) workshop at NeurIPS 2026 in Sydney will focus on discrete diffusion and flow models for next-generation decoding.

The event aims to move beyond the limitations of traditional autoregressive models—which generate output sequentially in a fixed order—by exploring iterative denoising for parallel, non-causal, and controllable generation. The workshop invites submissions on theories, algorithms, and applications, with a paper deadline of August 29, 2026.

Related event: NeurIPS 2026 Workshops to Explore Discrete Diffusion(2 posts)→

Original post →

More from Research

Research channel →