Kamijo and colleagues built a pipeline that denoises 4D-STEM scans without clean training images and then sorts crystal components automatically. It is aimed at beam-sensitive materials such as polymers, whose low-signal diffraction data are especially hard to analyse.
Reality
- Evidence50
- Adoption
- Insufficient
- Hype gap+15
- Incentives15
- Confidence45
CMAC researchers trained a crystallisation emulator on more than 20,000 simulated experiments and ran 360 optimisation campaigns on it. The rankings describe a model of one generic process, so process teams get a cheap first comparison of strategies before committing reactor time.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+20
- Incentives40
- Confidence50
A machine-learning-accelerated simulation in npj Computational Materials puts electron transfer in TiSe2/HfSe2 under 50 femtoseconds and nearly stops it in one stacking of ZrTe2/HfTe2, for under a tenth of the usual compute.
Reality
- Evidence52
- Adoption
- Insufficient
- Hype gap+28
- Incentives44
- Confidence55
Shi and Berkelbach evaluate higher-rung density functionals without self-consistency for orders of magnitude less cost, and their double-hybrid averages below 13 kJ per mol across 39 measured adsorption reactions.
Reality
- Evidence60
- Adoption
- Insufficient
- Hype gap+12
- Incentives22
- Confidence52
An ab initio study in npj Computational Materials traces the figure to symmetry-enforced phonon selection rules. The effective masses it computes are comparable to cubic diamond's, so mass cannot account for the gap.
Reality
- Evidence52
- Adoption
- Insufficient
- Hype gap+12
- Incentives38
- Confidence48
A team in Korea added orthogonal transformations to a generative model so fine-tuning would stop warping its latent space, and says the result becomes useful for hydrogen-evolution catalysts once transfer learning adds data.
Reality
- Evidence38
- Adoption
- Insufficient
- Hype gap+20
- Incentives30
- Confidence45
A pipeline described in npj Computational Materials screened more than 27,000 concrete papers and reports an extraction F1 as high as 0.98 across several models. Rice University has a provisional patent pending on the method.
Reality
- Evidence45
- Adoption
- Insufficient
- Hype gap+25
- Incentives65
- Confidence50
Overpotential is the part of a CO2 electrolyser's electricity bill that buys no product, and a computational screen from IIT Gandhinagar says three five-metal boride sheets may not need any of it, though those sheets so far exist only in the calculation.
Reality
- Evidence46
- Adoption
- Insufficient
- Hype gap+24
- Incentives72
- Confidence54
A fully time-resolved reconstruction of the defect network at a Sigma5(210) tilt boundary in aluminium finds migration carried by short-lived loops, which leaves the glide-mobility term in grain-boundary theory without an object to describe.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+12
- Incentives30
- Confidence52
A phase-field plus active learning loop inverted the process-microstructure problem for hard drive media, and beat expert best guesses. The abstract is short on how the win was measured.
Reality
- Evidence52
- Adoption
- Insufficient
- Hype gap+24
- Incentives58
- Confidence45
SeDeNet trains on noisy data alone and reports 95 percent or better parameter accuracy at one twentieth the exposure. That moves the photon budget into software, if you trust the output.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+28
- Incentives32
- Confidence44
A machine-learning potential for Co-Ni-V-H finds V-centered short-range order raises average hydrogen solution energies and thins out strong binding sites versus a random alloy.
Reality
- Evidence55
- Adoption
- Insufficient
- Hype gap+12
- Incentives22
- Confidence52
A published framework extracts an impurity Green's function on IBM processors using 8 qubits and 1 ancilla. That register size puts the near-term burden on circuit compression, not on qubit count.
Reality
- Evidence58
- Adoption12
- Hype gap+12
- Incentives30
- Confidence55