The vision-language system pairs 3D atomic arrangements with text descriptions and outperformed a text-only baseline on materials-selection tasks, including a search for solid-state battery electrolytes.
Research
Google pledges $3 million in compute credits for European researchers
Snorkel expands its open AI benchmark commitment to $30 million
OpenProblemBench turns to AI judges for unsolved math and physics
AgentHorizon puts AI judges to the near-miss test
Research
Epoch’s automation tests find research judgment lagging behind coding skill
Six models attempted the organization’s work without intervention. The strongest handled well-defined tasks, but a flawed pilot experiment exposed a harder problem.
Research
DOE plans up to $100 million for Ph.D. fellowship pilot
The program targets about 100 fellows on four-year pathways, with research placements at a national laboratory or user facility and in industry.
Reasoning-trace endpoints can match full-trace training, researchers say
Devin adds Claude Haiku 5.5 as a Fusion sidekick
PIR probe aims to tell hidden model knowledge from missing knowledge
UniMate paper proposes one text-to-motion model for any 3D skeleton
MIT researchers' InstructMesh tool aims to fix flaws in AI-generated 3D designs before printing
SimTrace paper proposes synthetic clickstreams to train simulated web users
ArgGYM proposes an engine-scored benchmark for defeasible reasoning
AIM paper proposes a manager for research ideas in automated experiments
More research news
Older stories →Q&A: The “collision” between AI, your brain, and society ↗
Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio ↗
Plan or Retrieve First? A Scale Study of Tool Discovery Architectures ↗
Research@ 2026: 10 ways we're amplifying human ingenuity ↗
Researchers improve weather forecasts looking two to six weeks out ↗
How artificial neural networks build their own learning machinery ↗
How Molmo 2’s open release helped an MIT researcher rethink fish tracking ↗
Q&A: Building accountability behind our words — and AI’s — in civic conversations ↗
Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings ↗
Google tests location embeddings to fill gaps in public-health models
Ai2 releases open AstaBrief model for cited scientific reports
arXiv caps submitters at two monthly papers as submissions reach a record
Researchers test inference-time method for modeling diverse preferences
MetaPersona draws on social-science studies to build simulated populations
Study finds aligned training data can prompt misaligned behavior in other contexts
Hermes-Learn trains models to adapt context use during reasoning