Planet AI Developer Filter
A preprint based on 646 adults compares AI models on video interviews, reporting a modest multimodal edge and more reliable ratings for targeted questions.
Planet AI Developer Filter
A preprint reports optical recovery of model and inference-state data from FPGA memory and compute boundaries, including results under incomplete readout.
Planet AI Developer Filter
A preprint reports global existence and uniqueness for a two-dimensional stress-diffusive Oldroyd-B model under specific mathematical assumptions.
Planet AI Developer Filter
A preprint reports a video model that learns which moments and tracked entities support driving-risk predictions from coarse video labels.
Planet AI Developer Filter
A preprint reports lower HPC job-runtime error and a 17-fold speedup over AVX; its end-to-end result comes from a tested ALCF benchmark configuration.
Planet AI Developer Filter
A simulated robot study found listwise visual feedback competitive with pairwise methods, though task and seed variation kept the result uncertain.
Planet AI Developer Filter
A theoretical model predicts optical peaks and a tunable nonlinear Hall response as a Hopf-link semimetal changes between linked and unlinked phases.
Planet AI Developer Filter
A preprint examines narrow orbital equilibria, relativistic precession, Lidov-Kozai cycles and S-cluster orientations around a spinning black hole.
Planet AI Developer Filter
A 2026 arXiv preprint reports higher average scores for sparse model merging, while noting added training overhead and mixed task results.
Planet AI Developer Filter
A preprint reports strong low-data results for speech-act-informed models forecasting online conversation derailment, but mixed rankings across domains.
Planet AI Developer Filter
A preprint proposes six threat-actor attributes to make pre-release misuse evaluations of open-weight frontier AI more consistent.
Planet AI Developer Filter
A preprint reports one proposed Vision MoE backdoor with under 9% dormant success in small batches and 76% to 87% activation success at batch size 128.