3 papers
cs.LG2026
DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models
Amin Karimi Monsefi, Dominic Culver, Nikhil Bhendawade +4
Diffusion large language models are a compelling alternative to autoregressive models, yet existing RL methods for diffusion treat all denoising steps as equally important and rely…
cs.LG2026
The Design Space of Tri-Modal Masked Diffusion Models
Louis Bethune, Victor Turrisi, Bruno Kacper Mlodozeniec +21
Discrete diffusion models have emerged as strong alternatives to autoregressive language models, with recent work initializing and fine-tuning a base unimodal model for bimodal gen…
q-bio.NC2025
Attention when you need
Lokesh Boominathan, Yizhou Chen, Matthew McGinley +1
Being attentive to task-relevant features can improve task performance, but paying attention comes with its own metabolic cost. Therefore, strategic allocation of attention is cruc…