Pinned Tweet
Can a static image editor learn the laws of physics? Yes, by learning from video dynamics! 🎥 Super excited to share our latest work on physics-aware image editing. Full story below!
Editing images is a series of state transitions between the source image and the edited image that we want. Yet, the existing paradigm doesn't explicitly include any transitioning priors in the editing process.
This becomes particularly prevalent for edits, involving causal dynamics (e.g., refraction, deformation).
To model this kind of physics-informed information, we leverage the rich priors present in videos and introduce PhysicEdit 🔥
TL;DR: We fine-tune QwenImage Edit on a curated dataset of videos with reasoning traces and fixed-length transition queries to do solid physics-aware image editing!
In the process, we introduce a cool dataset "PhysicTran38K", consisting of 38K transition trajectories across five physical domains and devise a method to provide supervision from it QwenImage Edit.
Hop in to learn more ⬇️
Liangbing Zhao retweeted
jev is insane 🤯
Here is Jev playing subway surfers at super human speed, and also playing 50 games at once.
cost less than a cent to do this run.
Jev does not replace llms like astra or fable, but opens up an entirely new world of capabilities.
Liangbing Zhao retweeted
GPT6 Astra solved Rubik’s Cube with robot hands
dex-rubik-cube.yanjieze.com/
Liangbing Zhao retweeted
Introducing H3 Max Camera Controls, the world's fastest 3D video model.
Turn a single viewpoint into a navigable 3D scene in less than 3 seconds.
Set the horizontal and vertical camera angle in degrees, and H3 Max Camera Controls moves through the scene while preserving geometry, position, and materials across perspectives.
Fast enough to control the camera in real time.
Liangbing Zhao retweeted
You can kind of just train specialized computer-use agents.
Here is a small Qwen3-VL-2B fine-tuned to do instance segmentation of cells through computer-use actions in a headless labeling app. It takes in screenshots as input and outputs commands to draw polygons, zoom, navigate, etc...
All navigation and labeling in this video is done by Qwen.
Inspired by Astra's computer use abilities, who can also label data quite well, albeit quite slowly.
Liangbing Zhao retweeted
🚀 Sol-H3: @MiniMax_AI H3 Video Generation Faster Than Playback 🤩
Five seconds of world. 1.653 seconds to infer.
We’re releasing Sol-H3, our fastest end-to-end MiniMax-H3 inference stack yet. On one 8× NVIDIA B300 Blackwell system, it generates five seconds of 1344×768 video with stereo audio in 1.653 seconds. Across 1×, 4×, and 8× B300, Sol-H3 reaches up to a 15.54× speedup versus Base H3.
Compared with 50-step Base H3 Dense on the same 8× B300 system, the four-step Sol-H3 profile delivers:
• 5s: 18.250s → 1.653s (11.04×)
• 10s: 50.660s → 3.732s (13.57×)
• 15s: 99.513s → 6.612s (15.05×)
Sol-H3 also scales across GPU counts:
• 4× B300: 2.918s / 6.993s / 12.542s for 5s / 10s / 15s (12.11–15.54×)
• 1× B300: 13.745s / 37.813s / 52.260s for 5s / 10s / 15s (9.45–14.29×)
All figures are medians of three measured runs after one warmup at 1344×768 and 24 FPS with stereo audio. Base H3 uses 50 scheduler points (49 DiT forwards); Sol-H3 uses four DiT forwards, so this is a full-profile comparison—not an attention-only runtime change. Sol-H3 uses Dense attention on 1× B300 and SOL with INT8 QKV / FP8 output transport on 4× / 8×. Timing includes text encoding, DiT denoising, and video/audio VAE decoding; model loading, compilation warmup, and final MP4 encoding are excluded.
Sol-H3 brings Sol-Engine × Sol-Attn into one full-stack runtime:
• dynamic sparse attention with no retraining
• fused norm, RoPE, MLP, and sparse-attention setup
• fused INT8 QKV / FP8 output communication across 8 GPUs
• parallel, batched VAE decoding
• precomputed AdaLN caching
Inside the stack:
• sparse-attention setup: 1.206 → 0.285 ms (−76.4%)
• VAE decode: 7.55 → 0.602 s
• ~24 GB memory freed per GPU
Any MiniMax-H3 few-step LoRA can plug into the same engine, and the code is deployment-friendly under Apache 2.0.
For us, the bigger milestone is crossing from “fast generation” into “faster than playback.” That opens the path toward continuous 24 FPS generation and truly interactive video systems.
We’re excited to partner with @reactorworld to release Sol-H3 and make it available as an API day-0.
Try it now on Reactor: reactor.inc/sandbox?model=fa…
🔗 nvlabs.github.io/Sana/Sol-En…
Amazing team effort—full credits in the blog.
@shanasaimoe @lawrence_cjs @yitongli165665 @haopengl33 @HaochengXiUCB @songhan_mit
Liangbing Zhao retweeted
Heard that some frontier models are basically a 48-layer transformer looped twice (48L x 2).
Now we are introducing DeepLoop: Depth Scaling for Looped Transformers (arxiv.org/abs/2607.13491),
making the loop transformer stable and scalable!
Liangbing Zhao retweeted
send me inputs and i will rotate them with Atlas. single image or sparse multiview.
Liangbing Zhao retweeted
Today, we're sharing new research on Solaris, our first Interface World Model.
Solaris is a new kind of operating system that generates interactive interfaces frame by frame, in real time, with no code. We find that Solaris outperforms frontier LLMs when generating new interfaces, across structural similarity and information retention. Read more and request early access at the link below.
Liangbing Zhao retweeted
Introducing Atlas:
The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D.
Model the world, move the camera, and simulate space & time.
Liangbing Zhao retweeted
😅😅😅
Never sample your diffusion noises in bf16/fp16. I spent my last two days to understand why my sampled images grey out.
Also AI is not good at these model training stuffs, this is what happened when you use AI to train a diffusion model.
Liangbing Zhao retweeted
🚢 Marin 535B-A23B started training this week! As usual, the whole process is open.
Voyage plan: pretraining (80%) + midtraining (20%) on 18.75T tokens on 11 x GB200 NVL72 for ~3 months (2.7e24 FLOPs). Post-training will follow.
Before kicking off the run, we trained a 4-rung scaling ladder from 1.6B-A61M (48B tokens) to 27.7B-A1.2B (926B tokens) to debug issues, and to make a forecast of our hero run. This is by far our biggest run, so definitely expecting the unexpected.
Liangbing Zhao retweeted
Introducing GEN-1.5, a one-shot learner.
It can learn new tasks in a few seconds. Show it what to do, and it generalizes.
This capability emerged from pretraining on physical data at scale, as a step towards our mission of building general intelligence for the physical world.
Liangbing Zhao retweeted
We’re introducing imagination models: a new foundation model architecture that unlocks learning from internet-scale video.
Our first imagination model, Photon-1, learned to use a computer by watching 18 years of screen recording video without action labels.
Liangbing Zhao retweeted
A new KAUST-led approach helps #AI understand physical change, improving image edits involving light, motion and materials.
Read more: cemse.kaust.edu.sa/articles/…
Liangbing Zhao retweeted
Editing images is a series of state transitions between the source image and the edited image that we want. Yet, the existing paradigm doesn't explicitly include any transitioning priors in the editing process.
This becomes particularly prevalent for edits, involving causal dynamics (e.g., refraction, deformation).
To model this kind of physics-informed information, we leverage the rich priors present in videos and introduce PhysicEdit 🔥
TL;DR: We fine-tune QwenImage Edit on a curated dataset of videos with reasoning traces and fixed-length transition queries to do solid physics-aware image editing!
In the process, we introduce a cool dataset "PhysicTran38K", consisting of 38K transition trajectories across five physical domains and devise a method to provide supervision from it QwenImage Edit.
Hop in to learn more ⬇️
Liangbing Zhao retweeted
if you want to learn about how we trained KREA Flux, we prepared a detailed blog in the link below:
krea.ai/blog/flux-krea-open-…
