researcher @ bytedance seed
Hong Kong
Joined February 2020
- Tweets163
- Following586
- Followers2.2K
- Likes1.3K
Pinned Tweet
Introducing Omni, one unified model can support any-to-any multimodal modeling, including multimodal understanding, image/video generation and editing, world modeling and 3D reconstruction. All in one that adopts standard mixture-of-experts arch with only 3B activations.
Multimodal understanding capabilities will significantly accelerate robotics one day, just like what happened before in visual generation.
impressive improvements, well done bro
🚀 LingBot-VA 2.0 is here!
After half a year of teamwork, we're thrilled to release LingBot-VA 2.0 — a native video-action foundation model for generalizable robot control.
Unlike prior world-action models that retrofit generic video generators for robot control, LingBot-VA 2.0 is natively pretrained from scratch as a video-action foundation model.
Three key insights:
🌍 Native Video-Action Pretraining for learning world knowledge that enables strong generalization.
🧩 Semantic Visual-Action Tokenizer for more accurate action prediction with robust prompt following
⚡ Foresight Reasoning enables the robot to think ahead while acting, delivering continuous, responsive control without interrupting execution.
It runs in real time on consumer-grade GPUs, supports up to 150 Hz control, and generalizes to unseen tasks.
👇 Demos below
Ceyuan Yang retweeted
Introducing EdgeBench, a benchmark designed to study how agents learn from environments over at least 12~72-hour runs. We find that performance follows a log-sigmoid function of environment interaction time with high precision.
EdgeBench is built with three ingredients:
- 🌍 Real & Diverse: 134 real-world tasks across 6 task categories, spanning scientific problems, professional knowledge work, software engineering, optimization, formal math, and games.
- ⏳ Ultra-Long-Horizon: Each task supports 12–72 hours of agent work. Recorded human effort averages 57.2 hours.
- 🔁 Informative Feedback: Agents receive real-world feedback for continuous improvement.
After 38,000 hours of agent runs on EdgeBench, a scaling law for learning from environments emerges:
- 📈 As agents interact with task environments over time, their aggregate performance is precisely fit by a log-sigmoid function.
- 🧠 This phenomenon can be explained by an elegant theory of graph exploration.
We are releasing an initial 51 of the 134 tasks, together with the full evaluation framework, to help advance long-horizon agent research. Check our blog & paper for more findings!
Blog edge-bench.org/
Paper edge-bench.org/paper.pdf
GitHub github.com/ByteDance-Seed/Ed…
Dataset huggingface.co/datasets/Byte…
Details below 👇🧵
see you in seoul
Adversarial Flow Models is accepted at ICML.
Continuous Adversarial Flow Models is accepted at ECCV.
Heading to Seoul next week. See you all at #ICML.
Ceyuan Yang retweeted
We are thrilled to introduce our new work, Representation Forcing for Bottleneck-Free Unified Multimodal Models!
What lies between language and pixels?
Representation.
We propose to ground visual representation in the decoder of a unified model. The decoder learns to predict it from text, just as the encoder reads it from images. Understanding and generation then meet in a single representation space, learned end to end, with no frozen VAE in between.
We call this Representation Forcing: the encoder's representations force the decoder to learn the same visual structure, and the decoder's predictions, in turn, force the pixels to follow it.
Paper: arxiv.org/abs/2605.31604
Project page: yuqingwang1029.github.io/Rep…
Huggingface: huggingface.co/papers/2605.3…
Feel free to chat with @yuqingwang1029 at #CVPR2026
Started last summer, but shared much later and less detailed than we hoped. It was built by a small team under many constraints, with plenty of imperfections, but also with a lot of support from people who truely believed in it. It only made us believe more: the future is multimodal.
Ceyuan Yang retweeted
Context Unrolling in Omni Models
- A unified multimodal model natively trained on diverse modalities, including text, images, videos, 3D geometry, and hidden representations
- Enables Context Unrolling, where the model explicitly reasons across multiple modal representations before producing predictions
- Achieves strong performance on both multimodal generation and understanding benchmarks, while demonstrating advanced multimodal reasoning capabilities, including in-context generation of text, image, video, and 3D geometry
Ceyuan Yang retweeted
Continuous Adversarial Flow Models (CAFMs)
Paper: arxiv.org/abs/2604.11521
Flow matching generates poor samples without guidance because the MSE loss induces incorrect generalization. Instead of an isotropic Euclidean distance, we need a manifold-aware criterion—but how can we obtain it?
CAFMs bring adversarial training to continuous time. Learning velocity with a discriminator induces better generalization because the discriminator as a criterion can learn the manifold!
Also unlike flow matching’s forward KL objective, adversarial training allows optimizing different divergences. CAFMs can generate sharper and higher-quality samples.
Adversarial training in continuous time also avoids the vanishing gradient problem, leading to stable training.
CAFMs can be trained from scratch or used to post-train existing flow models. Post-training SiT/JiT for just 10 epochs yields large FID improvements. We also observe significant GenEval and DPG improvements when post-training text-to-image models.
More details in this thread!
OpenSourcing week for Lingbot. Well done, bro
🚀 We’re excited to announce LingBot-VA, a new state-of-the-art robot policy model from
@robbyant_brain !
LingBot-VA is built on a causal, autoregressive video-action world model for generalist robot control.
Highlights:
(1) First unified autoregressive video-action world model for robot control
(2) Low-latency inference with a new asynchronous execution pipeline
(3) SOTA on RoboTwin (92.9%, firstever > 90%) and LIBERO (98.5%)
(4) +20% over π0.5 on challenging real-world long-horizon & high-precision tasks
Ceyuan Yang retweeted
End-to-End Training for Autoregressive Video Diffusion via Self-Resampling
huggingface.co/papers/2512.1…
Ceyuan Yang retweeted
Our research: Adversarial Flow Models (AF)
arxiv.org/abs/2511.22475
AF unifies Adversarial and Flow Models.
Unlike GANs, AF learns optimal transport (stable). Unlike CMs, AF only trains on needed timesteps (save capacity). We can train super-deep 112-layer 1NFE model! SOTA FIDs!
new one
Seedream 4.0 is the new leading image model across both the Artificial Analysis Text to Image and Image Editing Arena, surpassing Google's Gemini 2.5 Flash (Nano-Banana), across both!
Seedream 4.0 is the latest release from Bytedance Seed, and is a substantial improvement on their Seedream 3.0 model with improved text rendering performance. It is the new leader in the Text to Image leaderboard, and achieves parity with Gemini 2.5 Flash in Image Editing.
Seedream 4.0 maintains the same pricing as Seedream and SeedEdit 3.0 at $30 per 1k generations, and is currently available on @FAL , @replicate , and @BytePlusGlobal
See the thread below to see Seedream 4.0 in the Artificial Analysis Image Arena for yourself!
Ceyuan Yang retweeted
How do we generate videos on the scale of minutes, without drifting or forgetting about the historical context?
We introduce Mixture of Contexts. Every minute-long video below is the direct output of our model in a single pass, with no post-processing, stitching, or editing.
1/4
Long video generation usually results in context increasing/scaling during chunk/frame-wise rollout.
Considering context scaling may require context selection, we thus introduce the idea of MoE into long context modelling and propose Mixture of Contexts. All previous context/memory is considered while the chosen ones are computed in a data-driven manner. You can easily enjoy 7x compute savings.
Home page: primecai.github.io/moc/
ArXiv: arxiv.org/abs/2508.21058
Thank all coauthors for making this happen, especially to Shengqu @prime_cai. Somehow our work is also ROUTED by arxiv to graphics rather than CV :)
