[NeurIPS 2025] Official repository of the paper "Unlocking Aha Moments via Reinforcement Learning: Advancing Collaborative Visual Comprehension and Generation"
-
Updated
Sep 27, 2025 - Python
[NeurIPS 2025] Official repository of the paper "Unlocking Aha Moments via Reinforcement Learning: Advancing Collaborative Visual Comprehension and Generation"
DART (Diffusion-Autoregressive Recursive Transformer) is a novel hybrid architecture that combines diffusion-based and autoregressive approaches for text generation.
To associate your repository with the autogressive topic, visit your repo's landing page and select "manage topics."