eFinder

eFinder

Home › Institutional Prestige of Berkeley Lab › Article

MatterChat model helps AI to 'see' the language of atom-scale physics to sharpen materials predictions

Phys · · 1223 words · By Linda Vu
Institutional Prestige of Berkeley Lab AI Innovation in Physical Sciences Computational Efficiency

Researchers at Lawrence Berkeley National Laboratory have developed MatterChat, an AI framework that connects Large Language Models with physics-based models to improve predictions of material properties. The system uses a 'bridge model' to translate atomic-scale data into a format LLMs can process, outperforming general-purpose AI in specific materials science tasks.

open_in_new Read the original article: https://phys.org/news/2026-05-matterchat-ai-language-atom-scale.html

analyticsAnalysis

20%
Propaganda Score
confidence: 95%
Minor concerns. Some persuasive language detected, but largely factual.

psychologyDetected Techniques

warning
Loaded Language 80% confidence
Using words with strong emotional connotations to influence an audience.
warning
Glittering Generalities 70% confidence
Using vague, emotionally appealing phrases ('freedom', 'justice') without specifics.

fact_checkFact-Check Results

9 claims extracted and verified against multiple sources including cross-references, web search, and Wikipedia.

info Single Source 3
check_circle Corroborated 2
help Insufficient Evidence 2
verified Verified 1
verified Verified By Reference 1
check_circle
“a new AI framework from Lawrence Berkeley National Laboratory (Berkeley Lab), called MatterChat, solves this problem by creating a specialized "bridge."”
CORROBORATED
Multiple independent web sources confirm that Lawrence Berkeley National Laboratory developed MatterChat as a 'bridge' model connecting LLMs with physics-based AI.
menu_book
wikipedia NEUTRAL — Lawrence Berkeley National Laboratory (LBNL, Berkeley Lab) is a federally funded research and development center in the hills of Berkeley, California, and Oakland, California, United States. Establish…
https://en.wikipedia.org/wiki/Lawrence_Berkeley_National_Lab…
menu_book
wikipedia NEUTRAL — The National Energy Research Scientific Computing Center (NERSC) is a high-performance computing (supercomputer) research facility that was founded in 1974. The National User Facility is operated by L…
https://en.wikipedia.org/wiki/National_Energy_Research_Scien…
menu_book
wikipedia NEUTRAL — The University of California, Berkeley (UC Berkeley, Berkeley, Cal, or California) is a public land-grant research university in the Southside and Northside neighborhoods of Berkeley, California, Unit…
https://en.wikipedia.org/wiki/University_of_California,_Berk…
+ 3 more evidence sources
check_circle
“The resulting system already significantly outperforms general-purpose AI tools like GPT-4 at predicting material properties”
CORROBORATED
Multiple sources, including a research paper and news reports, state that MatterChat outperforms state-of-the-art LLMs (like GPT-4) in predicting material properties.
travel_explore
web search NEUTRAL — The results show that MatterChat consistently outperformed its competitors across a range of tasks. The model was more accurate in classifying material types and demonstrated superior precision in pre…
https://newscenter.lbl.gov/2026/05/18/new-matterchat-model-h…
travel_explore
web search NEUTRAL — Fig. 2 MatterChat accurately predicts material properties and outperforms state-of-the-art LLMs. (a) Illustration of multi-modal material property queries using MatterChat.
https://arxiv.org/pdf/2502.13107
travel_explore
web search NEUTRAL — Fig. 2 MatterChat accurately predicts material properties accurately and outperforms the state-of-art. LLMs. (a) Illustration of multi-modal material property queries using MatterChat.
https://www.researchgate.net/publication/389130005_MatterCha…
verified
“A paper describing this work was recently published in Nature Machine Intelligence.”
VERIFIED
Newswise explicitly states that a paper describing this work was published in Nature Machine Intelligence, and the journal itself is a verified entity via Wikipedia.
menu_book
wikipedia NEUTRAL — An AI boom is a period of rapid growth in the field of artificial intelligence (AI). The most recent boom happened in the early 2020s before seeing increased acceleration and media coverage. Examples …
https://en.wikipedia.org/wiki/AI_boom
menu_book
wikipedia NEUTRAL — Artificial intelligence is the capability of computational systems to perform tasks that are typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and…
https://en.wikipedia.org/wiki/Applications_of_artificial_int…
menu_book
wikipedia NEUTRAL — Artificial intelligence (AI) is the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and dec…
https://en.wikipedia.org/wiki/Artificial_intelligence
+ 3 more evidence sources
info
“Yingheng Tang, a postdoctoral researcher in Berkeley Lab's Applied Math and Computational Research Division (AMCR) and lead author on the paper.”
SINGLE SOURCE
The provided evidence for this claim consists of irrelevant search results regarding the Qin Dynasty, providing no information about Yingheng Tang.
travel_explore
web search NEUTRAL — The Qin dynasty (/ tʃɪn / CHIN[3]) was the first imperial dynasty of China. It is named for its progenitor state of Qin, a fief of the confederal Zhou dynasty (c. 1046 –256 BC).
https://en.wikipedia.org/wiki/Qin_dynasty
travel_explore
web search NEUTRAL — The Qin Dynasty (221–206 BC) reunited China and laid the foundation for 21 centuries of imperial rule. Its great building projects and achievements were overshadowed by enormous cultural destruction a…
https://www.chinahighlights.com/travelguide/china-history/th…
travel_explore
web search NEUTRAL — After centuries of wars between rival states, China entered a new era of unification under one empire beginning in the Qin and Han Dynasties. During this period, the emperor had supreme power over a c…
https://www.chnmuseum.cn/portals/0/web/zt/gudai/en/detail4.h…
info
“the team trained their bridge model on a dataset curated by pairing nearly 143,000 stable atomic structures from the Materials Project with their corresponding physical properties.”
SINGLE SOURCE
While the evidence confirms the use of the Materials Project API and the existence of the bridge model, the specific number '143,000' is not explicitly corroborated across multiple independent sources in the provided text, though it is mentioned in the context of the project's training data.
travel_explore
web search NEUTRAL — This training data was automatically assembled using the Materials Project’s API and deliberately enriched with properties fundamental to microelectronics design — like formation energy and bandgap — …
https://newscenter.lbl.gov/2026/05/18/new-matterchat-model-h…
travel_explore
web search NEUTRAL — Materials Project.
https://next-gen.materialsproject.org/materials/mp-6930
travel_explore
web search NEUTRAL — MatterChat consists of three core components: the Material Processing Branch, the Language Processing Branch, and the Bridge Model. The Material Processing Branch extracts atomic-level embeddings from…
https://mlciv.com/papers/tang2025matterchat.pdf
help
“the researchers benchmarked MatterChat against a suite of other AI systems, from general-purpose LLMs to other specialized scientific AI methods. The results show that MatterChat consistently outperformed its competitors across a range of tasks.”
INSUFFICIENT EVIDENCE
No evidence was provided for this specific claim in the search results.
info
“In a collaboration with Fermilab, MatterChat is already contributing to a U.S. Department of Energy Genesis Mission project—called Accelerating eXtreme Environment Specs-to-Silicon (AXESS)”
SINGLE SOURCE
The provided evidence for this claim consists of irrelevant search results regarding HTTP cache directives, providing no information about Fermilab or the AXESS project.
travel_explore
web search NEUTRAL — The no-cache directive in a response indicates that the response must not be used to serve a subsequent request i.e. the cache must not display a response that has this directive set in the header but…
https://stackoverflow.com/questions/866822/why-both-no-cache…
travel_explore
web search NEUTRAL — Cache directive "no-cache" An explaination of the HTTP Cache-Control header The Cache-Control header is used to specify directives for caching mechanisms in both HTTP requests and responses. A typical…
https://no-cache.net/
travel_explore
web search NEUTRAL — The .nocache.js file contains JavaScript code that resolves the Deferred Binding configurations (such as browser detection, for instance) and then uses a lookup table generated by the GWT Compiler to …
https://support.google.com/code/answer/77858?hl=en
verified
“The team also credits supercomputing resources at the National Energy Research Scientific Computing Center (NERSC)... access to the Perlmutter supercomputer”
VERIFIED BY REFERENCE
Wikipedia and official NERSC reports confirm that the Perlmutter supercomputer is located at NERSC, which is operated by Lawrence Berkeley National Laboratory.
travel_explore
web search NEUTRAL — Shyh Wang Hall, houses the National Energy Research Scientific Computing Center at Lawrence Berkeley National Laboratory.
https://en.wikipedia.org/wiki/National_Energy_Research_Scien…
travel_explore
web search NEUTRAL — The National Energy Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory (Berkeley Lab) has unveiled the first phase of its next-generation supercomputer, Perlmutter.
https://www.scientific-computing.com/news/perlmutter-superco…
travel_explore
web search NEUTRAL — The National Energy Research Scientific Computing Center (NERSC) at Lawrence Berkeley National Laboratory (Berkeley Lab) today formally unveiled the first phase of its next-generation supercomputer, P…
https://www.nersc.gov/news-and-events/news/berkeley-lab-depl…
help
“Yingheng Tang et al, A multimodal large language model for materials science, Nature Machine Intelligence (2026). DOI: 10.1038/s42256-026-01214-y”
INSUFFICIENT EVIDENCE
No evidence was provided for this specific citation in the search results.
info Disclaimer: This analysis is generated by AI and should be used as a starting point for critical thinking, not as definitive truth. Claims are verified against publicly available sources. Always consult the original article and additional sources for complete context.