Ph.D. Student at University of Cambridge. Building generative models for the real world.
Cambridge, UK
Joined September 2013
- Tweets13
- Following145
- Followers110
- Likes77
Pinned Tweet
Introducing S2PD: Serial-to-Parallel Diffusion.
Bidirectional video models fail to follow simple rules even when trained on enormous amounts of data. We show that serial computation is critical for generating consistent video and use it in the high-noise regime of video diffusion.
Project Page: jefequien.github.io/S2PD/
Jeffrey Hu retweeted
Working on Mirror convinced me of something too opinionated for a paper: we've been over-obsessed with "action" in embodied AI. (1/15)
Jeffrey Hu retweeted
Video generation can produce visually plausible clips while still violating basic physical laws or symbolic rules.
S2PD—Serial-to-Parallel Diffusion—addresses this by combining two stages: autoregressive diffusion at high noise to coordinate dependent events and valid state transitions, followed by parallel diffusion at low noise to refine the full video more efficiently.
Across games, physical simulations, and real video, S2PD follows rules more reliably than matched bidirectional baselines while offering greater temporal stability and improved sampling efficiency compared with other serial approaches. The method is presented with both a pixel-space diffusion transformer and a LoRA-adapted pretrained video model.
Title: S2PD: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation
Project URL: jefequien.github.io/S2PD/
Paper: arxiv.org/abs/2610.06847v1
Jeffrey Hu retweeted
S2PD: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation
TL;DR: S2PD combines autoregressive diffusion at high noise with parallel diffusion at low noise, improving rule-following, physical consistency, temporal stability, and sampling efficiency in video generation.
arxiv.org/abs/2610.06847
All code, datasets, and models have been released.
github.com/jefequien/S2PD
Also, check out The Serial Scaling Hypothesis (arxiv.org/abs/2507.12549) and The Seriality Gap (arxiv.org/abs/2607.13031), which reach similar conclusions about the importance of serial computation in video generation.
Jeffrey Hu retweeted
ruilong has been one of the best folks in the world in speeding things up. when he releases something, it's gonna be both performant and easy-to-use.
World models are moving beyond offline generation towards interactive, real-time experiences.
Introducing ⚡FlashDreams⚡: an open-source high-performance inference and serving library built for autoregressive world models:
🔥 Up to 3.10× faster LingBot-World inference
🔥 Up to 2.12× faster Self-Forcing inference
🔥 Up to 1.40× faster Wan2.1 inference
🔥 8 integrated models
🔥 Multi-GPU, streaming, low-latency serving
🔥 Agentic skills that teach you how to use it
FlashDreams is designed for a new generation of AI systems that continuously evolve over time while responding to user interactions. It powers applications across robotics, autonomous vehicle simulation, gaming, and virtual worlds.
Github: github.com/NVIDIA/flashdream…
Docs: nvidia.github.io/flashdreams
Research page: research.nvidia.com/labs/sil…
Join the #flashdreams Discord channel at discord.gg/yTdHDqFP
FlashDreams is also the runtime backbone behind NVIDIA OmniDreams (github.com/nv-tlabs/omni-dre…)
1/n
#AI #WorldModels #FastInference #PhysicalAI #OpenSource #NVIDIA
Jeffrey Hu retweeted
[SIGGRAPH Asia '25] Editable Physically-based Reflections in Raytraced Gaussian Radiance Fields
Contributions:
• A reconstruction method for radiance fields, with distinct optimization for diffuse and specular components, using path tracing for the latter.
• An efficient and accurate training method that reconstructs the diffuse and specular components of the scene in a single representation.
• An efficient ray tracer for Gaussian primitives, fast enough to enable treatment of multiple bounces with minimal computational overhead.