Han Lin
555
posts
@hanlin_hl
PhD
@UNC| MS
@Columbia| Research Intern
@AIatMeta| Working on generative models, multimodal learning, and LLMs
Chapel Hill, NC
Joined December 2021
Pinned
🚀 Excited to share V-Co, a diffusion model that jointly denoises pixels and pretrained semantic features (e.g., DINO). We find a simple but effective recipe: 1️⃣ architecture matters a lot --> fully dual-stream JiT 2️⃣ CFG needs a better unconditional branch -->
Glad to share that V-Co is accepted to #ECCV2026! 🇸🇪 ✨ V-Co introduces a practical representation learning recipe for image generation by jointly denoising pixels and pretrained semantic features (e.g., DINOv2). Its design combines a fully dual-stream JiT architecture,
🚀 Excited to share V-Co, a diffusion model that jointly denoises pixels and pretrained semantic features (e.g., DINO). We find a simple but effective recipe: 1️⃣ architecture matters a lot --> fully dual-stream JiT 2️⃣ CFG needs a better unconditional branch -->
Spatial reasoning is not just about getting the answer right — it is also about knowing when the image does not provide enough evidence to answer, and what additional viewpoint is needed to resolve the uncertainty. ✨Excited to share SpatialUncertain, a controlled framework for
🚨 Excited to share SpatialUncertain — a controlled framework for evaluating whether VLMs know when not to answer spatial questions (and why). ➡️ Spatial reasoning is not just about finding the right answer—it is about knowing whether the available evidence supports an answer at
🌟Glad to introduce PhyMotion: a structured 3D motion reward for physics-grounded human video generation. Realistic human motion remains a major challenge in video generation. Existing rewards often stay in 2D pixel space, missing failures such as floating feet,
🚨 Excited to introduce PhyMotion🤸: Structured 3D Motion Reward for Physics-Grounded Human Video Generation! ❌ Existing 2D video rewards misleadingly assign high scores to videos with floating feet, self-penetrating limbs, and physics-violating motions. ✅ PhyMotion lifts
