Nature Communications
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 3
- Reconfigurable oxide device powers reservoir-computing digit demo
According to cnBeta’s report on research published in Nature Communications, University of Würzburg researchers developed an oxide-interface device that can switch among transistor, memristor and memcapacitor modes. Built from lanthanum aluminum oxide and strontium titanate, the device supported a reservoir-computing circuit that recognized digits from 5×4 black-and-white images. In the reported test configuration, the device-based circuit consumed about 2.71 nJ per input pulse, compared with about 16.57 nJ for a transistor-and-memristor control circuit.
cnBeta · 🔥 9 - AI-Generated Fake Papers Tested Against IEEE Access Defenses
IEEE Spectrum documented an experiment in which free versions of ChatGPT and Gemini generated a paper describing research that was never performed, while Analemma’s FARS system produced a second paper describing AI-conducted computational research. IEEE Access screened both papers with the publisher’s cooperation; neither was accepted, peer reviewed, or independently validated, though the papers exposed flaws such as fabricated citations and nonsensical figure artifacts. The report also cited estimates that detectable LLM influence reached 57% of academic articles in 2025, up from 12% in 2023, while emphasizing that detectable AI use does not by itself establish low-quality “AI slop.”
IEEE Spectrum · 🔥 0 - MIT researchers create multimodal AI model for crystalline materials
MIT researchers have created a multimodal AI model for crystalline materials that combines 3D atomic structures with natural-language descriptions, producing unified accounts of chemical composition, structure, properties, processing considerations and potential applications. In a Nature Communications paper, the team reported that the model, trained with contrastive learning on AI-generated crystal descriptions, identified better material designs than models trained on text streams alone and learned geometric representations that text-only models did not. In an example, its top candidates for solid-state lithium-ion battery electrolytes included materials independently identified as promising; the researchers compare the approach to AlphaFold.
MIT News · AI · 🔥 0
Experience and discussion from the community
Share my Nature Communications experienceAsk about Nature Communications
Nobody has shared their experience with Nature Communications yet.