Hanyu Wang (@hywang66) header
Hanyu Wang (@hywang66) profile picture

Research Scientist at Luma AI.

Joined December 2021
Photo by Hanyu Wang (@hywang66)Photo by Hanyu Wang (@hywang66)Photo by Hanyu Wang (@hywang66)Photo by Hanyu Wang (@hywang66)Photo by Hanyu Wang (@hywang66)
Watching a robot finally succeed after hours of trial and error is so satisfying. In SEAR, agents wrote code, tested it on real hardware, and kept revising it from experience. I had a lot of fun building this loop 🤖🦾 Paper’s out! arxiv.org/abs/2610.18792
An LLM robot agent’s success rate goes up as it practices. Did it learn anything, or did it just try more often? And do the skills and memories it saves still help on new tasks, in a new simulator, on a real robot? We built SEAR to answer both. To our knowledge, it’s the first large scale controlled study of self-evolving robot agents. The paper is now on arXiv 📄 The agent improves a robot in one of three ways: it writes code that becomes the policy, it acts live in the control loop, or it trains its own VLA. 7 frontier LLMs, 330 sim tasks, 5 real-robot tasks, 4+ years of agent runtime, with a matched control for every comparison and a sealed judge on unseen scenes. A few findings surprised us 🧵 1/7 arxiv.org/abs/2610.18792
1
4
148
Moving the camera nearly halves VLA success, while adding noise to the robot’s proprioceptive input barely affects it. Which setup changes matter most? The differences might surprise you. Introducing LIBERO-RIG, our new benchmark for VLA robustness: arxiv.org/abs/2610.18843
A VLA can ace LIBERO and still fail when the camera moves slightly. Excited to share our new work, LIBERO-RIG! We keep tasks fixed and perturb 11 aspects of the robot hardware setup. Across 8 fully evaluated VLAs, pooled success drops from 93.2% to 50.9%. arxiv.org/abs/2610.18843
2
102
Had a lot of fun leading the design and experiments for SEAR’s self-evolving harness on real robots 🤖! Coding agents can get more done on real robots than you’d think, and surprisingly fast ⚡️. Check out sear.bot!
Amazed by Astra controlling robots? It turns out other LLMs can do it too – even the text-only ones like DeepSeek. We also find Astra did best watching and correcting every move live; Fable did best building its own tools and policy in code first. We introduce SEAR, a new benchmark and framework to rigorously evaluate and study how LLM agents self evolve on robots. 🔁📈The loop: within a controlled time budget, an agent and its library update a policy → the policy runs in simulation or on hardware → a judge returns feedback → the library carries the skills and lessons into the next task.  💻🚏 Self-evolving agents operate in myriad ways - we study all of them at once: the policy can be code the agent wrote (code-as-policy), the agent itself in the control loop (agent-as-policy), or a VLA it trained (autoresearch). 🧑‍🔬🤖We spent 4+ years of accumulated agent evolving time across simulation and real hardware and tested 7 frontier models (GPT-6-Astra, Fable 5.1, Opus 5, GPT-5.6-sol, Gemini-3.6-flash, Kimi-k3, DeepSeek-v4-flash) on 330 sim tasks, 2 physics engines, and real robots with different embodiments.  🌐Project overview: sear.bot  A thread 🧵1/5
1
1
12
349
Hanyu Wang retweeted
Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects you're working on self-driving cars. All in the name of safety, of course. Because malicious actors controlling the world’s operating systems, inboxes and cars would be extremely dangerous!
mythos will be bad ON PURPOSE on ai "frontier llm research" tasks, this is very very sad for the research community also the fact that this is un purpose not visible to the user is crazy
94
744
51
6,658
369,251
Hanyu Wang retweeted
Exciting news: UNI-1.1-Max and UNI-1.1 debuts making @LumaLabsAI the #3 lab in the Image Arena across both Text-to-Image and Image Edit! These are versions released without agentic search. Text-to-Image Arena - UNI-1.1-Max #6 overall (1193), +12 points over MAI-Image-2 - UNI-1.1 #7 overall (1190), +13 points over Reve-v1.5 Multi-Image Edit Arena - UNI-1.1-Max #7 overall (1315), +21 points over Seedream 4.5 - UNI-1.1 #8 overall, (1298) Single-Image Edit Arena - UNI-1.1-Max #7 overall (1337) - UNI-1.1 #11 overall, (1310) on par with Grok-Imagine-Image (20260207) Congratulations to @LumaLabsAI on this solid performance!
The Uni-1.1 API is live today. Built-in prompt enhancement, research, and reference gathering at the API level. Trained in collaboration with Hollywood cinematographers, VFX artists, and world-class artists across cultural forms. Less than half the price and latency of comparable models. Designed for builders shipping in production — and ranked top 3 lab in the Image Arena across Text-to-Image and Image Edit. Start Building → lumalabs.ai/api
18
34
14
246
118,103
Hanyu Wang retweeted
#3 on Image Edit. #3 on Text-to-Image. @arena The compute we did it with would surprise you. Proud of this team @LumaLabsAI ! (A slightly more detailed report will come out soon)
Exciting news: UNI-1.1-Max and UNI-1.1 debuts making @LumaLabsAI the #3 lab in the Image Arena across both Text-to-Image and Image Edit! These are versions released without agentic search. Text-to-Image Arena - UNI-1.1-Max #6 overall (1193), +12 points over MAI-Image-2 - UNI-1.1 #7 overall (1190), +13 points over Reve-v1.5 Multi-Image Edit Arena - UNI-1.1-Max #7 overall (1315), +21 points over Seedream 4.5 - UNI-1.1 #8 overall, (1298) Single-Image Edit Arena - UNI-1.1-Max #7 overall (1337) - UNI-1.1 #11 overall, (1310) on par with Grok-Imagine-Image (20260207) Congratulations to @LumaLabsAI on this solid performance!
2
17
153
28,544
Hanyu Wang retweeted
Most image models are good at one thing. Uni-1 has been good at everything we've thrown at it. Our team generated thousands of images leading up to Uni-1 launch. We embedded them all into a single map where visual similarity determines proximity. The result speaks for itself.
22
44
6
322
49,768
Hanyu Wang retweeted
We are loving the energy around Uni-1! Quick note since we’re seeing questions: With Luma Agents, requests can route across models. If you want to make sure you’re using Uni-1, here’s how: - Select Create Image → Uni-1 - Or, explicitly ask the agent to use Uni-1 - Check the model label on outputs to confirm API access coming soon for more direct testing. Keep the feedback coming, and keep on creating → lumalabs.ai/uni-1.
19
45
6
251
20,189
Hanyu Wang retweeted
Uni-1 is here! A new kind of model that thinks and generates pixels simultaneously. Less artificial. More intelligent.
449
757
689
4,730
6,515,422
Hanyu Wang retweeted
Introducing Uni-1, Luma’s first unified understanding and generation model, our next step on the path towards unified general intelligence. lumalabs.ai/uni-1
31
97
27
781
232,992
Hanyu Wang retweeted
Excited to introduce Uni-1, our new *unified* multimodal model that does both understanding and generation: lumalabs.ai/uni-1 TLDR: I think Uni-1 @LumaLabsAI is > GPT Image 1.5 in many cases, and toe-to-toe with Nano Banana Pro/2. (showcase below)
29
54
14
411
99,964
Hanyu Wang retweeted
Introducing Luma Agents. Creative agents that make you prolific. You set the direction. They build with you, seeing what you see and helping teams explore further, iterate faster, and watch ideas multiply.
52
107
51
969
7,992,085
Hanyu Wang retweeted
Excited to introduce Uni-1, our new multimodal model that *unifies* understanding and generation. TLDR: a team of ~15 researchers is going pound-for-pound with nano banana and gpt image 🧵
Excited to introduce Uni-1, our new *unified* multimodal model that does both understanding and generation: lumalabs.ai/uni-1 TLDR: I think Uni-1 @LumaLabsAI is > GPT Image 1.5 in many cases, and toe-to-toe with Nano Banana Pro/2. (showcase below)
22
64
9
535
78,857
We’re excited to announce UPLiFT, our lightweight, pixel-dense feature upsampler. UPLiFT boosts feature density, preserves semantics, and has better efficiency scaling than recent SOTA methods. See all links in the thread below. Coauthors: @_sakshams_ @AnirudAgg @abhi2610 🧵[1/6]
8
51
3
388
19,409
Hanyu Wang retweeted
🎉 Excited to share our paper "Trokens: Semantic-Aware Relational Trajectory Tokens for Few-Shot Action Recognition" has been accepted to #ICCV2025! Equally co-led with @ShuaiyiH — we advance few-shot action recognition via smart point tracking. 🔗 trokens-iccv25.github.io 🧵👇
6
26
2
143
10,820
📢📢 Excited to announce that LARP has been accepted to #ICLR2025 !🎉 Congrats to my co-authors and see you in Singapore🇸🇬!
🚀 Introducing LARP: Tokenizing Videos with a Learned Autoregressive Generative Prior! 🌟 📄 Paper: arxiv.org/abs/2410.21264 📜 Project page: hywang66.github.io/larp/inde… 🔗 Code: github.com/hywang66/LARP Collaborators: @_sakshams_ , Yixuan Ren, @HaoChen_UMD , @abhi2610 #GenAI
1
2
14
1,115
I will be presenting LARP at ICLR today. 🎤 Oral: 11:18 AM – 11:30 AM (UTC+8), Oral Session 3C 🖼️ Poster: 3:00 PM – 5:30 PM (UTC+8), Hall 3 + Hall 2B, Poster #162 You’re very welcome to drop by for discussion and feedback!
1
1
5
646
🌟 CoLLM: A Large Language Model for Composed Image Retrieval (CVPR 2025) ✨A cutting-edge training paradigm using image-caption pairs 📊High-quality synthetic triplets for training & benchmarking 🔗Project: collm-cvpr25.github.io/ 📄Paper: arxiv.org/abs/2503.19910 #LLM #CIR
2
7
3
22
4,820
Hanyu Wang retweeted
I saw a slide circulating on social media last night while working on a deadline. I didn’t comment immediately because I wanted to understand the full context before speaking. After learning more, I feel compelled to address what I witnessed during an invited talk at NeurIPS 2024 by Professor Rosalind Picard. I deeply respect Professor Picard’s scholarship and contributions to the field. However, her comments during the talk reflected a deeply troubling and racist view of Chinese scholars. This was not just inappropriate but also profoundly disheartening. First, it was entirely unnecessary to mention the student’s nationality when discussing an incident of cheating. The point about academic integrity could have been made without emphasizing nationality. Yet, Professor Picard chose to highlight it. This choice perpetuates harmful stereotypes about Chinese scholars and reflects a broader bias against Asians, often rooted in the assumption that we “work hard, avoid conflict, and don’t push back.” This needs to change. Asians, like everyone else, have the right to speak out and demand accountability when racism occurs. We will ensure that being racist against Asians has consequences, including here, Professor Picard. What made this incident worse was how it unfolded during the Q&A session. A Chinese attendee asked a professional and thoughtfully articulated question. She began by thanking Professor Picard for her talk and posed this question: Are you calling out the student’s nationality because you find most Chinese scholars honest, and the fact that the cheating student was Chinese is rare? Is that why you emphasized nationality? This was a generous and high-EQ question, offering Professor Picard an opportunity to reconsider or clarify her comments. Unfortunately, she doubled down instead. Professor Picard reinforced her remarks by quoting the student’s excuse —that ethics wasn’t taught in their school—and generalized this as a broader issue with Chinese education. This statement is both factually incorrect and deeply offensive. There are glaring logical flaws in this argument: 1.If the student cheated, why would their excuse about ethics education be taken at face value? A serious scholar would investigate the claim before making it a central part of their argument. 2.Even if the student’s school didn’t teach ethics (which is false for schools in China), other sources like family and community often instill strong ethical values. Ignoring this nuance is careless and reinforces stereotypes. What is most heartbreaking is that Professor Picard couldn’t even acknowledge something as simple as: “Most Chinese scholars are honest and upright.” Instead, she focused on the singular exception and added, “Of course, with this one exception in this case” in her response. I regret that this happened at NeurIPS. I regret that this happened in my research community—a place I have cherished and contributed to for over 14 years. I regret that this happened at MIT, an institution of excellence and aspiration for many Chinese scholars. Racism has no place in academia, and incidents like this tarnish the principles of inclusion and respect that we, as a global research community, should uphold. I hope NeurIPS and the broader academic community take this as a wake-up call to address the biases and systemic issues that enable such comments to go unchallenged. We must do better. @MIT_CSAIL @NeurIPSConf
It is just so sad that the #NeurIPS2024 main conference ended with such a racist remark by a faculty when talking about ethics. How ironic! I also want to commend the Chinese student who spoke up right on spot. She was respectful, decent, and courageous. Her response was exemplary: she began by acknowledging the speaker’s efforts, then gave the speaker an opportunity to clarify (though, regrettably, the speaker’s reply only reinforced her bias), and ultimately called attention to the inappropriate racial bias and offered constructive suggestions. Thank you for speaking out!
49
173
9
1,033
156,630
Mitigating racial bias from LLMs is a lot easier than removing it from humans! Can’t believe this happened at the best AI conference @NeurIPSConf We have ethical reviews for authors, but missed it for invited speakers? 😡
172
772
337
3,690
2,229,115