Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers Mohamed Bahgat, Steven R Wilson, and Walid Magdy introduced Lexpansion, a supervised learning approach that uses dictionary definitions as input features to a BERT transformer model to expand lexicons for soc…
Researchers Max Brede, Felix Schönbrodt, Birk Hagemeyer, and Veronika Lerche introduced the Automated Motive Coder (AMC), a machine learning tool that automates coding of Picture Story Exercise (PSE) narratives with accu…
Researchers from Apple and academic institutions introduced Dual-Stage Registers (DSR), a register-based intervention that reduces outlier tokens in Diffusion Transformers (DiTs) for image generation, improving generatio…
A developer demonstrated a local AI voice agent running on a $50 Arduino Uno, using a distilled transformer quantized to 4-bit or binary weights to fit within the ATmega328P's 32KB RAM and 256KB flash. The system process…
A Databricks-focused consultancy, Zephico, breaks down the real costs of migrating from Snowflake to Databricks, arguing that the true benefits come from moving for ML and streaming workloads rather than just cost saving…
TensorFlow Recommenders (TFRS) with a two-tower retrieval model is the recommended open-source starting point for video recommendation in an OTT platform, according to a technical response. The response suggests using TF…
A new tutorial from StarTrail-org introduces PixelRAG, a pixel-native retrieval-augmented generation pipeline that renders web pages and PDFs as images, divides them into overlapping tiles, and generates multimodal embed…
A developer has created the first known project integrating Meta's smart glasses with fingerspelling translation software, using a neural network trained on Google's FSboard dataset with a CNN + GRU temporal encoder and …
Pinterest is hiring a Staff Data Scientist for its Notifications team, offering a salary of $165k–339k per year for a remote position based in San Francisco, CA, US. The role involves developing measurement frameworks, i…
Mistral released Shieldstral, a 3B-parameter multimodal safety classifier under Apache 2.0, that accepts plain-language moderation policies at inference time and returns a calibrated score from a single output token, run…
A developer shared open-source experiments on integer-only LLM inference, including a Q16.48 fixed-point implementation for a tiny character GPT and TinyLlama-1.1B, and a mixed F11/F12 TinyLlama candidate that matched al…
Roboflow published a tutorial on detecting small objects in drone imagery, training RF-DETR on roughly 7,000 aerial images with over 120,000 person annotations to identify nine classes of people and vehicles. The workflo…
Pinterest is hiring a Senior Manager of Machine Learning Engineering-Applied Research in San Francisco or remote, offering $228k–469k per year, to lead a team of ML researchers and engineers building next-generation reco…
A new wave of AI hardware and software co-design is replacing heavy floating-point matrix multiplications with addition-only, ternary weight sets {-1, 0, 1}, enabling up to a 32-fold reduction in weight memory and a 5.3-…
EdotEnv, a Y Combinator S26 company, has launched reinforcement learning environments that use quantitative trading workflows to teach large language models research methodologies. The platform generates trading tasks wi…
MagicLAG, a maker, developed a method to strengthen 3D-printed parts by embedding carbon-fiber cloth between a core and outer shell layers, more than tripling the yield strength of test hooks compared to standard prints.…
Roboflow published a tutorial on building dwell time and zone analytics with vision AI, using RF-DETR for person detection, ByteTrack for persistent tracking, and a zone timer to measure per-person dwell time and unique …
A new study from arXiv presents the first production-scale characterization of agentic coding workloads using GitHub Copilot traces from June 2026, covering 3.2 million users, 13 million sessions, 761 million LLM calls, …
Inflect-Micro-v2, a complete local text-to-waveform speech synthesis model with under 10M parameters, has been released, offering fixed-voice English TTS with deterministic seeds, long-text handling, and CPU or CUDA infe…
Artificial intelligence now powers over 80% of digital advertising decisions, with the AI-driven ad tech market projected to reach $370 billion by 2028 and 40% of ad creative AI-assisted. The systems use collaborative fi…