AI for Science

A study of the Station environment finds that multi-agent systems can recover a substantial share of known scientific findings without web access or revealed results. Yet sustained exploration should not be confused with mature scientific judgment.

OpenAI has released nearly 400 AI-generated mathematical results across more than 700 manuscripts. Researchers are impressed by the scale, but warn that understanding and verifying the material could take years.

Nikon has revoked the first-place award in its Small World in Motion contest after determining that the winning microscopy video violated rules covering generative AI.

Gan Jiang is a self-learning agent for powder X-ray diffraction that improves executable analysis skills without retraining its language model. It links phase identification, multiphase decomposition, and physics-constrained whole-pattern modelling.

Researchers used a language-model agent to modify the Sella molecular geometry optimizer, with safeguards against premature stopping and overfitting. The resulting AutoSella variants reduced force calls on held-out molecules and transferred to DFT tests.

SciUtopia is a persistent closed-loop simulation in which LLM agents choose research topics, collaborate, submit papers, review, seek funding, and evolve across simulated years. Its experiments suggest that more submissions can raise output while also increasing review pressure and reducing researcher retention.

A team at Vals.ai used AI agents to design and screen magnetic materials, producing one new candidate and resurfacing another first synthesized in 1999. Both combine semiconductor gaps with zero net magnetization and predicted spin selectivity, although substantial experimental work remains.

Fold2Reason turns protein structures into verifiable reasoning supervision and tests whether non-linguistic scientific data can transfer beyond biology. The reported results show gains across structure, spatial, graph, scientific, and general reasoning tasks.

Rules to Tools converts written scientific requirements into callable executable checks for coding agents. The study finds higher repair scores on some SciCode cohorts, but the gains are task-dependent and can come with higher CPU usage.

OSWorld-Science brings scientific workflows into computer-use evaluation, covering molecular drawing, pathology analysis, statistics, and physical simulation. Instead of judging clicks alone, it examines whether agents produce verifiable scientific artifacts.

Automated research is not only about running more experiments; it also requires deciding which research directions deserve attention. AIM makes idea organization, selection, implementation auditing, and budget allocation explicit parts of the search loop.

EvoDuet treats web retrieval as an evolving part of scientific problem solving rather than a fixed add-on. Its experiments show meaningful gains across optimization tasks, although the benefits vary substantially by model.

Large language models are becoming better at proving difficult mathematical statements, but proving a theorem does not automatically make it interesting. A new study proposes a measurable way to rank mathematical discoveries and build a self-expanding formal library.

Anthropic says Claude identified a previously uncharacterized enzyme system linked to repeated DNA sequences in bacteriophages. The system, called array-associated reverse transcriptases, or ARTs, may be interesting—but its biological function remains unknown.

SAIR Foundation has announced an Open Math Model Initiative to develop open-weight models and supporting tools for mathematical research. Its first phase will target proof understanding, literature checking, coding, and formalization.

Anthropic has confirmed that it operates a wet biology lab in the Bay Area, giving its models a path from generating hypotheses to testing them in physical experiments. The move expands the company’s role from AI provider to participant in scientific validation.

Flash floods can develop faster than conventional monitoring systems can respond. TACLS combines atmospheric moisture data from GNSS signals with machine learning to help forecasters identify rising flood risks earlier.

HypoEvolve turns collaboration among scientific LLM agents into an explicit evolutionary process. Specialized agents propose, challenge, revise, and retain hypotheses, which are evaluated in a cancer drug-repurposing setting against external biological evidence.

ScienceIDE addresses the scientific experience bottleneck by converting fragmented research repositories into executable and verifiable environments for AI agents. The team also trained the PhAI-IDE model family from verified interaction trajectories.

OpenAI says its systems made major progress on the Navier–Stokes problem, one of the Millennium Prize Problems. The achievement has instead triggered debate over data, authorship, collaboration, and the growing race to dominate mathematical research.

According to the source material, 25 Fields Medalists have endorsed a statement warning that some AI companies are reducing mathematical research to benchmark performance. Their concern is not the use of AI itself, but the preservation of understanding, attribution, education, and open collaboration.