No Channels Left Behind: Understanding Diffusion Acceleration through Residual Gates
Paper • 2610.40914 • Published • 21
Dongyoon Han
calintz
AI & ML interests
Interested in efficient modeling - VMs, VLMs, LLMs, and pretty much all things in machine learning
Recent Activity
authored a paper 7 days ago
No Channels Left Behind: Understanding Diffusion Acceleration through Residual Gates authored a paper 7 days ago
LaLa-World: Scaling Latent World Models with Natural Language Actions authored a paper 7 days ago
HumanSkill-4D: Training-Free Language Querying of Animatable Gaussian Avatars in Space and TimeOrganizations
authored 3
papers 7 days ago
upvoted a paper 7 days ago
upvoted a paper 8 days ago
upvoted a paper 18 days ago
authored 11
papers 24 days ago
Speculative Forcing for Efficient Autoregressive Video Generation
Paper • 2610.22347 • Published • 79
Omni-Modal LLMs Know More Than They Say
Paper • 2610.16135 • Published • 12
Learning Transferable World-Action Models from Task-Paired Human Videos
Paper • 2610.14351 • Published • 12
On-Policy Distillation with Negative-Policy Rollouts
Paper • 2610.07874 • Published • 20
DLoop: Looped Speculative Decoding
Paper • 2610.07659 • Published • 14
Learning to See What You Need: Gaze Attention for Multimodal Large Language Models
Paper • 2605.13080 • Published
Revisiting Complete Reasoning Traces for Post-Training
Paper • 2609.07103 • Published • 18
Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs
Paper • 2609.04753 • Published • 15
Verification-Aware Training for Speculative Decoding
Paper • 2608.30135 • Published • 11
On-Policy Delta Distillation for Multilingual Math Reasoning
Paper • 2608.05802 • Published • 11
Retrieve, Don't Retrain: Extending Vision Language Action Models to New Tasks at Test Time
Paper • 2606.15631 • Published • 17
upvoted a paper 25 days ago
authored a
paper 4 months ago
upvoted a paper 4 months ago