Zhenze Yang (@yang_zhenze) header
Zhenze Yang (@yang_zhenze) profile picture

LLM, AI for science Currently at @periodiclabs, ex-xAI, ex-ByteDance, ex-MIT

Joined April 2020
Photo by Zhenze Yang (@yang_zhenze)
🚀
We built high-throughput materials labs in Menlo Park to create a loop between experiments and models. The labs generate fresh data, the models learn from it, and then help us decide what to try next. Using only 1,300 H200s, plus months of our experimental data, we mid-trained and RL’d an open-source model to surpass GPT-6 Astra on our analysis benchmark. We call it Neon. This is real footage from our lab. We’re focusing first on hard problems in materials science, including superconductors, magnets, and semiconductor materials. Read our blog posts below.
1
6
458
Zhenze Yang retweeted
We built high-throughput materials labs in Menlo Park to create a loop between experiments and models. The labs generate fresh data, the models learn from it, and then help us decide what to try next. Using only 1,300 H200s, plus months of our experimental data, we mid-trained and RL’d an open-source model to surpass GPT-6 Astra on our analysis benchmark. We call it Neon. This is real footage from our lab. We’re focusing first on hard problems in materials science, including superconductors, magnets, and semiconductor materials. Read our blog posts below.
274
528
289
5,090
1,676,545
Glad to see this is finally out!
Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency. x.ai/news/grok-4-5
12
727
An early beta of Grok Build, an agentic CLI for coding, building apps, and automating workflows is now available for SuperGrok Heavy subscribers. Through this early beta, we will improve the model and product based on your feedback. Try it at x.ai/cli
1,547
1,294
797
9,605
56,755,586
Nature has severely outpaced humans in developing #multifunctional, #hierarchical #materials that access impressive material properties, all the while being fully #degradable and part of native #ecosystems. But how can we effectively model the intricate time and length scales in biological systems to translate design principles to meet engineering demands? Multi-agent #GenerativeAI is enabling us to develop rigorous strategies to learn complex design principles from biology and translating them towards engineering design. In a new white paper co-authored with @rachelkluu, Sofia E. Arevalo, Wei Lu, Bo Ni, Zhenze Yang, Sabrina Shen, Jaime Berkovich, Yu-Chuan Hsu, Stone Zan, & Markus J. Buehler we discuss how this new technology empowers critical new strategies essential for addressing contemporary materials design challenges to address higher functionality, more diverse raw materials as input, and sophisticated target material properties. We also discuss societal and ethical implications of this new technology. "Learning from Nature to Achieve Material Sustainability: Generative AI for Rigorous Bio-inspired Materials Design": doi.org/10.21428/e4baedd9.33… The paper is part of MIT's open access collection of papers about Generative AI from interdisciplinary teams of faculty and researchers across the Institute to explore the transformative potential of generative AI. Our contribution was a wonderful collaboration with a selection of members from our lab @LAMM_MIT who each contributed their unique perspectives. Pushing the frontier  Human-AI collaborations or multi-agent systems show intriguing results that exceed those that have been generated by human intelligence alone. AI, engendered by its radically different modes of cognition, could push against the limitations of established thought and provide insight into an as-of-yet unimagined future. Future artificial general intelligence could come to be seen as another milestone in human evolution, helping us to overcome barriers thought to be impenetrable today. We postulate that generative artificial intelligence (AI) can play a crucial role in solving this interdisciplinary challenge, translating insights across disparate knowledge domains and forming the basis for a new sustainable materials economy. The emerging next generation of AI systems surpasses limitations imposed by original training data, actively exploring new understanding. Bio-inspired generative AI notably widens the design space, fostering natural scientific discovery while departing from cyclical human-centric design, pushing the frontier of biomateriomic research. Generative AI can not only address complex forward and inverse tasks but also develop ontological knowledge that offers interpretability using graph theory, facilitating the translation of knowledge across diverse scientific domains. #GenerativeAI #Science #Engineering @mitpress
5
1
24
4,640
Thanks @MaterHoriz for selecting our review paper "Artificial intelligence and machine learning in design of mechanical materials" in the Materials Horizons 10th anniversary regional spotlight collection: The Americas.
1
1
4
712
See the link to check the collection: pubs.rsc.org/ru/journals/art…
1
439
You can't necessarily judge a book by its cover, but you may now be able to do the equivalent for materials of all sorts, from an airplane part to a medical implant. In new research published in #AdvancedMaterials @AdvSciNews, we report a deep-learning system that explores materials’ interiors from the outside. By feeding only surface data, or incomplete information about a material's state, the model predicts detailed information about internal structures, stresses, voids, and cracks. The method has implications as a tool for nondestructive testing, materials analysis and design, and research tasks where data is scarce or we can only acquire partial measurements. Read more on this on MIT News: news.mit.edu/2023/deep-learn… Paper: lnkd.in/ekYMSKCd
1
3
1
31
6,204
After a 3 year hiatus we held our 2022 Lab Holiday Party last night. It was special that many of our earliest members of the @LAMM_MIT family (2006 onwrds) could attend & meet current members. It's a blessing to be able to work with so many amazing people @MIT. Thank you! #F22MRS
5
1
72
Biology provides many interesting material designs, & proteins in particular are rich in complex structure-function relationships. In our latest paper we develop a set of proteome-inspired molecular & architected materials w/ unique mechanical properties authors.elsevier.com/a/1fzyP…
1
5
23
Is the #GriffithTheory still applicable to #2Dmaterials? The authors review experimental and theoretical studies on #fracture of 2D materials as well as challenges and opportunities for this budding field. #MRSBulletin @ProfBuehlerMIT @JunlouRice mvnt.us/m1348318
What are the physical laws that govern fracture at the extreme limit when materials consist only of few atomic layers? We review progress in this field in a new paper with @JunlouRice in @MRSBulletin: rdcu.be/cV328
1
7
1
34
Excited to share our new work of implementing a graph neural network for translating structural defects to atomic properties for crystalline solids such as graphene and metals.
The rise of #nanotechnology promises atomistically precise materials, but we lack efficient tools to explore the huge design space. We solve this challenge with an end-to-end neural net to rapidly relate atomic designs to mesoscale properties. Paper: nature.com/articles/s41524-0…
1
3
How to prevent fracture? Excited to share new work: Hierarchical Multiresolution Design of Bioinspired Structural Composites Using Progressive Reinforcement Learning, just published in Adv. Theory. & Sim. @AdvSciNews Read the full paper here: doi.org/10.1002/adts.2022004…
1
6
1
42
In new research from our lab reported in the J. Appl. Phys., led by grad students Wei Lu & Zhenze Yang, we report a new computational method to create synthetic 3D spider webs using a combination of multiscale modeling, graph & transformer neural networks aip.scitation.org/doi/10.106…
2
13
2
58
Excited to share our new work for creating and characterizing graphene metamaterials based on high-throughput technique and machine learning. Thanks for the support from @ProfBuehlerMIT .
Great work by my student @yang_zhenze @mit_dmse @MITEngineering: A new method to design & characterize graphene metamaterials for applications in energy storage & electrochemical catalysis.
4
Can computers design materials from human language cues? Our paper in @FrontMaterials "Words to Matter: De novo Architected Materials Design Using Transformer Neural Nets" w/ @yang_zhenze explains how 🎬Transformer based mechanism design doi.org/10.3389/fmats.2021.7… @FrontiersIn
1
6
39
Next Monday, I will present at USNCCM16 about our work on using deep learning methods to predict strain and stress fields in composites.
AlphaFold2 is open-source. Together with another comparable work RoseTTAFold:github.com/RosettaCommons/Ro…
Last year we presented #AlphaFold v2 which predicts 3D structures of proteins down to atomic accuracy. Today we’re proud to share the methods in @Nature w/open source code. Excited to see the research this enables. More very soon! https://nitter.cf/t.co/6uiV51Xly5 https://nitter.cf/t.co/CLo7EKubBT
New AI tool calculates materials’ stress and strain based on photos: The advance could accelerate engineers’ design process by eliminating the need to solve complex equations. mitsha.re/jBTe50EIPKC
7
54
5
197