Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A paper submitted to arXiv on 20 Sep 2026 introduces ArtifactBench, a lineage-aware evaluation suite for AI-generated music detectors that separates calibration from final testing and reports source-level performance wit…
A new investigation into video diffusion models finds that their tendency to generate content violating real-world physical laws stems from flaws in the models' internal attention mechanisms rather than from missing exte…
A new training method called Category-Aware Iterative Expert Training (One to More, More to One) targets uneven progress in repository-level software engineering agents, where pooled agentic reinforcement learning produc…
FloatLib, a verified arbitrary-precision floating-point arithmetic library written in Lean, has been released with support for IEEE binary and decimal formats, arbitrary-width posits, P3109, small ML formats, and user-de…
An engineer preparing optimizer experiments for an upcoming PyTorch Conference talk found that the Dion3 implementation in OLMo-core broke training with a learning rate death spiral while testing AdamW, Muon, and Dion. T…
A paper proposes a modular cognitive architecture in which neural computation is treated as only one of several possible substrates of cognition, with functions distributed among rules, explicit memory, specialized tools…
A developer built a fully local, browser-based sleep monitoring system that uses Whisper.cpp compiled to WebAssembly to filter out speech for privacy and a TensorFlow.js CNN to classify Mel Spectrograms of breathing soun…
A new tool called scfingerprint maps StarCraft: Brood War accounts to their human players by analyzing how they play, addressing a scene where progamers like Flash earned a $250,000 salary and over $700,000 in prize mone…
AMD researchers introduced Zebra-HyLo, an upcycling recipe that converts pretrained Transformer models into long-context hybrid LLMs through post-training on roughly 10B tokens, without pretraining from scratch. The meth…
A developer compared five approaches to a banking-support classification task using the BANKING77 dataset: a TF-IDF lexical model, a small sentence transformer (MiniLM), a general-purpose language model run through DSPy,…
A first-author paper presented at IEEE COMPSAC 2026 found that pull request acceptance can be predicted from submission-time signals alone, while review effort can only be predicted modestly, using the AIDev corpus of hu…
TypeSafe AI unveiled Jev, its first "System One model" that accepts text input but returns floating point numbers for categories, yes/no questions, ratings, and confidence scores instead of text. Jev charges only for inp…
Waymo posted a 2027 summer internship for MS and PhD students in software or machine learning engineering focused on simulation, based on-site in Mountain View, California. The role involves building multimodal data pipe…
Capsize Games released SpikeForge 0.4.0 and 0.5.0, splitting its Python spiking-neural-network toolkit into four separately shipped packages: the core spikeforge training workflow, spikeforge-targets conversions, the off…
A GitHub repository named moa-attention-verified-mullin, published under the account womenflyplanes, claims to be "the first attention kernel proven minimal BEFORE code was written," according to its README. The MIT-lice…
A developer maintaining trainproof, a deterministic linter for ML training runs, withdrew the strongest evidence behind the tool's validation gallery after two independent forensic audits found the cited log had been tru…
Roboflow is hiring a Research Scientist for its Roboflow Labs applied research unit in San Francisco at a salary of $250,000 to $450,000 per year plus equity, a range the job board says sits 49% above the $234,000 median…
Thinking Machines Lab posted a mid-training research role in San Francisco paying $350,000 to $475,000 per year, a range the job board says sits 76% above the $234,000 median for Core ML roles in the United States and hi…
Shrijith Venkatramana, an engineer building the LiveReview AI code review tool, published a technical explainer on LLM quantization showing that activation outliers concentrated in a handful of feature dimensions can con…
A developer modeled LLM dependence as an epidemic using the newly released epidemik 0.2.0, inspired by an arXiv preprint titled "Large-Language Models as a Cognitive Virus." The simplified SIR-style model splits a popula…