Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers David Ha and Jürgen Schmidhuber built the first AI to learn by rehearsing inside its own generated "dream" simulation, solving the CarRacing benchmark with a score of 906 against a 900 passing bar and survivi…
Yifan Zhang published a 2026 technical report describing the Recurrent Looped Transformer (RLT), an architecture pairing a causal encoder with a recurrent decoder that carries its final hidden state and layerwise sliding…
Engineering teams deploying reinforcement learning with verifiable rewards (RLVR) post-training have found that building the training harness accounts for only 20% of the work, with the remaining 80% going to verifier de…
Edge0-35B-A3B, a 35B-parameter sparse mixture-of-experts model built on Qwen3.5-MoE, loses an average of 3.9 points across five benchmarks when quantized to 4-bit through its edge0 inference pipeline versus its fp16 base…
A controlled test using a DDPM U-Net on CIFAR-10 found that LoRA rank 4 achieved the best FID score of 124.1380, slightly beating rank 8 at 124.2136, according to data from arXiv:2609.10656v1. The study, which tested ran…
A developer published a guide demonstrating how Scikit-Learn's Pipeline and ColumnTransformer modules can be combined into a reproducible machine learning workflow, using the classic Titanic survival dataset as a case st…
Shrijith Venkatramana, an engineer building the AI code review tool LiveReview, traces modern LLM tokenization back to Byte Pair Encoding, a 1994 data-compression algorithm by Philip Gage. The writeup explains how BPE-st…
Researchers Ilyas, Santurkar, Tsipras, Engstrom, Tran and Madry published a 2019 paper arguing that adversarial examples are not bugs but features of machine learning models, challenging the prevailing view that such per…
A paper submitted on 28 Sep 2025 to arXiv's Computer Science > Artificial Intelligence section formalizes the correspondence between fibring of modal logics and fibring of neural networks, a gap the authors say was never…
The Defense Innovation Unit has issued a solicitation, the Space Threat Intelligence Synthesis Engine, seeking AI software that can fuse live video, satellite imagery, radar, sensor feeds, geospatial data and classified …
A technical post argues that matrix calculus is unnecessary for differentiating functions of vector or matrix variables in machine learning, advocating instead for writing out tensor indices and applying ordinary differe…
The Alignment Research Engineer Accelerator (ARENA) hosts its full course content at arena.education, covering deep learning fundamentals, transformer interpretability, reinforcement learning, evaluations, and alignment …
Researchers published "A Mathematical Framework for Transformer Circuits," a paper that reverse engineers toy attention-only transformers with two layers or fewer and identifies "induction heads" as the mechanism explain…
A developer ran a controlled experiment with 40 agents (later replicated with another 40) to test whether the presence of a language model causes engineers to patch symptoms rather than trace root causes. Replacing the L…
Developer jens published a project called "Artificial Neural Networks - Sprite Layers" on June 20, 2025, a building kit that uses a single "Layer" sprite to construct a classical Rosenblatt perceptron, with duplicated La…
An INT8 quantization experiment on a TinyML ECG arrhythmia detection model cut its size from 87.5 KB in FP32 to 34.6 KB, a roughly 60% reduction, while reported accuracy held at 93.8%, according to the Beat2Bit project w…
A survey of Thai-language AI resources on Hugging Face highlights underused datasets and models, including the OpenThaiGPT evaluation dataset and the WangchanThaiInstruct multi-turn conversation set, alongside SCB 10X's …
A developer behind the Edge0 project demonstrated a 35B-parameter language model running on an iPhone with only 1–2.5 GB of peak memory by using Mixture of Experts (MoE) routing to load only the needed expert weights fro…
Google Research released TimesFM-3, a 330-million-parameter forecasting model that analyzes time series data alongside related information and known future events such as sales promotions or weather forecasts. Rather tha…
An independent developer has open-sourced Beatriz Epistemic Gate, a lightweight defensive proxy that detects data poisoning during LLM fine-tuning by verifying generated text against an immutable anchor corpus via embedd…