@Z1hanW

CS PhD student @berkeley_ai, 4D Vision + Robotics | MSCV @CMU_Robotics | Research Intern @ Frontier AI & Robotics (FAR)

z1hanw.com
Joined March 2022
  • Tweets122
  • Following298
  • Followers544
8Media
Humans generalize from rich interaction experience. We bring this to humanoid loco-manipulation through scalable real-to-sim-to-real. Introducing PRISM, we expand 4 real videos into 256 counterfactual variants, recover robot-object trajectories, and train one policy that generalizes across diverse objects, scales, and spatial layouts. CoRL’26 🚀 Website for details 👉 prism-real2sim2real.github.i… Work done at Amazon FAR w/ @zhenkirito123 @pabbeel @rocky_duan @JitendraMalikCV @carlo_sferrazza @ckarenliu @GuanyaShi @akanazawa
12
77
15
372
154,519
Zihan Wang retweeted
Amber Xie@amberxie_17h
Collecting data for dexterous hands is hard. Teleop needs a full setup. Bringing UMI-like devices to 20-dof hands is challenging. Introducing DexPort 🖐️: a hand-held interface w/ a real robot hand. No embodiment gap, recorded joint state AND actions, easy to use.
11
39
7
167
18,236
Zihan Wang retweeted
Jiaxin Ge
@jiaxin_ge_Oct 23
We trained an imaginative agent! ✨ Agima: an omni agent that plans by imagining futures. 𝗦𝗙𝗧: Directly generalizes to harder tasks! 𝟮%→𝟰𝟬%, beats GPT-5.6 Sol (𝟮𝟬%). 𝗥𝗟: New ways to imagine emerge! Imagines a second plan only if the first one fails (𝟯𝟮%→𝟭𝟬𝟬%).
16
59
9
353
43,685
Zihan Wang retweeted
C. Zhang@ChongZzZhangOct 23
With @JayHe748646 @KehanWen170077 @leggedrobotics , we push humanoid mobility to next frontier. Achieved by a generalist contact planner + a generalist tracker + a generalized neural mapping, warmstarted from few human data. Website: leggedrobotics.github.io/ACP… Promo Video 👇
26
70
18
460
42,226
Zihan Wang retweeted
Yuxuan Kuang
@yuxuank02Oct 22
What can a robot hand learn from seconds of human motion? It turns out: dexterous skills that generalize! Excited to share AdaDex: Learning Adaptable Dexterous Skills via Hybrid Dexterous Mimicking. From one short hand–object clip per skill, AdaDex learns an adaptable dexterous skill, with skills spanning functional grasping, object reposing, in-hand reorientation, and throw-and-catch. Project page: yuxuank.com/AdaDex/ 🧵👇 (1/8)
18
61
7
337
25,227
Zihan Wang retweeted
Qiayuan Liao
@qiayuanliaoOct 21
We found a way to scale humanoid control that reduces GPU-hours by more than 1000×. Across 129,514 motions, SuperMimic surpasses the state of the art (128 GPUs, 7 days) with 1 GPU in 3 hours. Project: supermimic.github.io Demo: supermimic.github.io/demo/
83
288
115
1,698
236,759
Zihan Wang retweeted
Yutang Lin
@yutang_llinOct 20
How can we represent whole-body interaction for humanoid robots—compactly enough for low-frequency planning, yet richly enough to preserve high-frequency feedback control? Introducing FIBER! 🤖 FIBER learns a unified latent interface between high-level planning and whole-body control, enabling a humanoid to acquire diverse interaction behaviors from heterogeneous experts through a pretrained, frozen Behavior Foundation Model (BFM). With a single planner and BFM, we demonstrate heavy-door pushing, box carrying, object grasping, controllable navigation, and more! 🌍 Website & Live Demo: fiberaa.github.io 📄 Paper: arxiv.org/abs/2610.24675 A thread on the ideas behind FIBER 🧵👇
4
21
3
118
5,822
Zihan Wang retweeted
Bowen Li@Bw_Li1024Oct 13
Failures are inevitable for robots that are deployed in the *big* world. We introduce ReSYNC, a system that leverages large-scale RL sim2real to recover from local failures and concept discovery to explain/avoid new failures. Check it out here: jaraxxus-me.github.io/ReSYNC…
12
28
4
85
7,941
Zihan Wang retweeted
Donghoon Ahn
@donghoon_ahnOct 13
Can diffusion models generate high-quality images without CFG? Yes! We introduce ✨Correction Training for Diffusion Models✨, which significantly improves sample quality without relying on classifier-free guidance. Project page: sunovivid.github.io/correcti… 🧵 (1/n)
36
142
29
1,129
154,096
Zihan Wang retweeted
Haven Feng
@HavenFengOct 13
Now autoregressive video models can remember. 🧠 Introducing MosaiChunk: persistent visual memory for long-horizon autoregressive video generation🧩. Instead of remembering everything, it remembers the right things. ✍️ 1/6
5
58
9
292
31,303
Zihan Wang retweeted
Yulin Liu
@yulinnliu07Oct 8
🤖 For parallel-jaw grippers, high-level plans can already be executed with mature primitives such as grasping. What about dexterous hands? Introducing GOTT (General Object Trajectory Tracking): object trajectories to dexterous execution with a reusable cross-embodiment contact primitive. The same primitive supports diverse real-world tasks including sweeping, pouring, whiteboard erasing, and hammering. With one checkpoint, GOTT reaches 90-96% grasp-and-lift success across multiple objects on three real-world arm-hand platforms. Check videos 👇
11
35
7
140
14,278
Practice makes perfect. A truly agentic robot shouldn’t just know what to do. It should also figure out how to get better. What if a robot could watch a demonstration, figure out what it needs to practice, and learn how to get better through practice? Introducing RPG: Guided Self-Improvement for Embodied Agents, a real-to-sim-to-real pipeline that turns real-world demonstrations into targeted practice. RPG reconstructs practice tasks from demonstrations, uses demo videos to guide skill improvement, and continually updates a shared skill system from failures. On real robots, RPG achieves 30/30 successes across 3 tasks. In simulation, it reaches 95.0% success across 22 manipulation tasks after 15 rounds of self-improvement. Demo → Reconstruct → Guided Practice → Self-Improve → Go Real
9
53
15
227
166,313
Zihan Wang retweeted
Chung Min Kim@ChungMinKimOct 6
Excited to introduce QF3: Fast Flow RL with Filtered Q-Gradients! Train humanoids from scratch, fine-tune VLAs, even fine-tune image models... all with one simple off-policy update ✨
13
59
7
424
39,330
Zihan Wang retweeted
Jiaheng Hu@JiahengHu1Oct 6
We’ve all seen the cool “Astra for robotics” demos. But behind the scenes? Slow rollouts, hand-engineered APIs, and/or simulated behaviors that struggle on real robots. How do we make coding agent actually work in the physical world? 🦾 Introducing SimEX: Simulation-Integrated Robotics AutoResearch 🧵 arxiv.org/abs/2609.38982
11
60
8
296
32,129
Zihan Wang retweeted
Sirui Xu
@xu_siruiOct 6
How can humanoids keep getting better on their own? Two loops: 🔁 Training: a data flywheel that grows from its own loco-manipulation skills: sirui-xu.github.io/InterMimi… 🧩 Test time: robots that behave better as an LLM agent thinks longer and evolves: sirui-xu.github.io/InterEvol… 🧵
9
31
1
136
7,022
Zihan Wang retweeted
Angjoo Kanazawa
@akanazawaOct 5
Since Skill from Video (2018) and Videomimic (2025) the unspoken challenge has been finding the right videos to teach robots at scale. Zihan's work changes that with v2v, turning a few good demos into many counterfactual videos. Another step towards scalable real2sim2real!
Zihan Wang
@Z1hanWSep 30
Humans generalize from rich interaction experience. We bring this to humanoid loco-manipulation through scalable real-to-sim-to-real. Introducing PRISM, we expand 4 real videos into 256 counterfactual variants, recover robot-object trajectories, and train one policy that generalizes across diverse objects, scales, and spatial layouts. CoRL’26 🚀 Website for details 👉 prism-real2sim2real.github.i… Work done at Amazon FAR w/ @zhenkirito123 @pabbeel @rocky_duan @JitendraMalikCV @carlo_sferrazza @ckarenliu @GuanyaShi @akanazawa
1
7
1
64
18,808
Zihan Wang retweeted
Jiaxin Ge
@jiaxin_ge_Oct 5
Generation 𝗖𝗔𝗡 improve understanding! With the right recipe, we show that training on generation data can help understanding tasks. Introducing OmniTaskonomy: our year-long study of when and how generation helps understanding.
13
58
10
299
61,224
Zihan Wang retweeted
C. Zhang@ChongZzZhangOct 2
video to video is a nice idea of data generation, which ensures both grounding and versatility.
Zihan Wang
@Z1hanWSep 30
Humans generalize from rich interaction experience. We bring this to humanoid loco-manipulation through scalable real-to-sim-to-real. Introducing PRISM, we expand 4 real videos into 256 counterfactual variants, recover robot-object trajectories, and train one policy that generalizes across diverse objects, scales, and spatial layouts. CoRL’26 🚀 Website for details 👉 prism-real2sim2real.github.i… Work done at Amazon FAR w/ @zhenkirito123 @pabbeel @rocky_duan @JitendraMalikCV @carlo_sferrazza @ckarenliu @GuanyaShi @akanazawa
5
54
6,444
Nice work, Zihan! Still remember the late nights before CoRL, you on PRISM and me on VLK, both doing real-to-sim-to-real. Glad to see PRISM released! 🚀🚀🚀
Zihan Wang
@Z1hanWSep 30
Humans generalize from rich interaction experience. We bring this to humanoid loco-manipulation through scalable real-to-sim-to-real. Introducing PRISM, we expand 4 real videos into 256 counterfactual variants, recover robot-object trajectories, and train one policy that generalizes across diverse objects, scales, and spatial layouts. CoRL’26 🚀 Website for details 👉 prism-real2sim2real.github.i… Work done at Amazon FAR w/ @zhenkirito123 @pabbeel @rocky_duan @JitendraMalikCV @carlo_sferrazza @ckarenliu @GuanyaShi @akanazawa
1
23
2,479
(3/4) Generated videos diversify object positions and poses. Generalization goes beyond new objects: trained only on flat terrain, the same policy picks up objects on a 35° ramp and from a 0.43 m elevated support in simulation, without any fine-tuning.
1
6
527
(4/4) We’ll release our code, model checkpoints, and all data. The release is currently under review. Stay tuned! More demos: prism-real2sim2real.github.i…

PRISM · Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation

A little human experience, amplified through counterfactual video generation.

prism-real2sim2real.github.io
7
437
Load more