AI that can actually "see" a crystal 🔬

Most of the solid stuff around us, from battery materials to chips, is made of crystals: atoms stacked in neat, repeating 3D patterns. Large language models are great at reading papers about those materials, but they have never really looked at the structures themselves. A team at MIT just showed a clever way around that, and the trick is one I keep coming back to: good synthetic data.

What they built

The model is called CLaC, short for Contrastive Language-Crystal model, and it is described in a paper in Nature Communications by Yang Jeong Park, Mayank Kumaran, Chia-Wei Hsu, Elsa Olivetti and Ju Li. MIT News covered it on 18 October 2026.

If you know CLIP, the model that taught image generators to connect pictures and words, you already get the idea. CLaC has two branches: a graph neural network that reads the 3D crystal structure, and a text encoder that reads a description. Contrastive training pulls matching structure-text pairs together in one shared space and pushes mismatched pairs apart.

The interesting part is where the text comes from. Real papers are a weak teacher here: they talk about a small set of popular "hot" materials and rarely spell out properties in a way a model can learn from. So the team used about 126,000 crystals from the Materials Project database, each paired with a narrative that a language model wrote from the computed properties: structure, behaviour, processing and possible applications. Those narratives cover roughly 300 times more different crystals than the literature text they compared against.

Why I like the result

The comparison is the honest part. They trained the same model twice: once on the synthetic narratives, once on text pulled from about four million academic papers.

  • In a zero-shot test on elemental composition, the synthetic version scored 95.4%, against 17.8% for a small text-only model (Gemma-3-1b-it).
  • When asked in plain language for materials with a target property, the paper-trained version did about as well as random picking. The synthetic version clearly did better.
  • Asked for lithium materials with a high band gap (a starting point for solid-state battery electrolytes), all of its top 30 results contained lithium. Two of them, Li6PClO5 and Li4SiO4, have been flagged or used as solid electrolytes in independent work.

So the lesson is not "more data". It is "clearer data". Text written to explain properties taught the model more than millions of real paragraphs did. That is the same pattern we keep seeing in LLM training and distillation.

Keeping it honest

The authors are open about the gaps. Telling crystal systems apart only reached about 34%, roughly on par with the small text-only baseline, which they link to small training batches and unbalanced data. Ionic conductivity, the property that really matters for electrolytes, isn't in the dataset at all. The model only handles single, perfect crystals for now, not defects, polycrystals or metal-organic frameworks. And synthetic text can hallucinate, so better ways to generate it are on their to-do list.

Still, Ju Li compares the direction to AlphaFold for proteins, and the open release makes it easy to build on: the code is on GitHub and the narrative dataset is on Hugging Face. Imagine typing "a stable, wide-band-gap lithium compound for a solid electrolyte" and getting a ranked shortlist of real structures. We aren't there yet, but this is a real step. ⚡

Sources

Reacties

Een reactie posten

Populaire posts van deze blog

SpaceX just asked to fly Starship hundreds of times a year 🚀

Last week the U.S. Federal Aviation Administration took a quiet but important step: it opened a formal review of SpaceX's request to launch Starship from Starbase in South Texas up to 456 times per year . Today the site is cleared for only 25 flights a year. That jump is the clearest signal yet that Starship is meant to become routine hardware, not a rare test rocket. What the FAA is reviewing On 14 October 2026 the FAA published a Notice of Intent to prepare a Programmatic Environmental Impact Statement (PEIS) for SpaceX's Starbase plans. A PEIS looks at a whole program of operations rather than a single launch, which is exactly what you need if you want a reusable rocket to fly the way airlines fly planes. SpaceX's proposal comes in two phases, according to the Federal Register notice: Phase 1: up to 100 launches and 200 landings per year (100 Ship, 100 Super Heavy), any day or time of week. Phase 2: up to 456 launches and 912 landings per year at Starbase. La...
Meer lezen

Google's Deep Review wants to be the second pair of eyes on every long document 🔍

AI has made writing fast. Checking has not kept up. That gap is exactly what Google Research is going after with Deep Review, which went public on 14 October. It's an agentic system that reads a long technical document and tries to find what's wrong with it: broken proofs, logic gaps, unsupported claims. For anyone working on legal and document AI, like me, this is a launch worth a close look. From conference pilot to public tool Deep Review is the grown-up version of Google's Paper Assistant Tool (PAT). Over the past year, authors of more than 16,000 manuscripts at STOC 2026, ICML 2026, NeurIPS 2026 and ICLR 2027 opted in for private feedback before the submission deadline, fully separate from official peer review. The need is real: Google counts over 74,000 submissions to ICLR, NeurIPS and ICML this year, up from about 23,000 in 2023. Now anyone can use it. Every user gets two free reviews per week, and you can connect a Gemini API key for higher limits and programmati...
Meer lezen