Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
JetSpec, a new speculative decoding method, trains a causal parallel draft head over fused hidden states from a frozen target model, enabling lossless verification of candidate trees in one forward pass. On Qwen3-8B with…
Researchers introduced JetSpec, a speculative decoding method that trains a causal parallel draft head on a frozen target model to draft entire speculative trees in one pass while preserving autoregressive factorization.…
Researchers introduced Self-Harness, a new paradigm enabling LLM-based agents to iteratively improve their own operating harnesses without human intervention. In tests on Terminal-Bench-2.0, Self-Harness boosted pass rat…
Roboflow launched RF-DETR Keypoint, a real-time end-to-end pose model that outperforms YOLO26-pose on accuracy and speed, with calibrated per-keypoint uncertainty and an Apache 2.0 license. The model is available as a pr…
Intel released Optimization Zone 1.1, a centralized repository of tuning guides for maximizing performance on Intel Xeon processors and some desktop CPUs. Newly covered workloads include Apache Cassandra with QuickAssist…
The field of artificial intelligence marks its 70th anniversary since being formally established at the 1956 Dartmouth Summer Research Project. AI has evolved from early neural network concepts by McCulloch and Pitts and…
Roboflow released a tutorial on building a planogram compliance detection workflow using its RF-DETR model and Google Gemini. The system automates shelf auditing by detecting empty zones and flagging brand mixing violati…
Roboflow released a tutorial on building an AI-powered robotic welding defect detection system using its RF-DETR model and Google's Gemini 2.5 Pro to classify welds as good or bad and generate automated inspection report…
Google Cloud announced the general availability of BigQuery Managed Python User-Defined Functions (UDFs), enabling data practitioners to run custom Python code directly within BigQuery using standard SQL queries or BigQu…
A new market simulation by Victor Haghani and Rich Dewey shows that blind extrapolation—investors naively assuming recent returns will continue—produces realistic market phenomena like excess volatility, bubbles, and cra…
Baidu released PP-OCRv6, a family of OCR models scaling from 1.5M to 34.5M parameters, supporting 50 languages. The medium model achieves 86.2% detection Hmean and 83.2% recognition accuracy, improving over PP-OCRv5_serv…
NLTK remains a viable tool for advanced text preprocessing despite the rise of LLMs. Three key tricks—MWETokenizer for preserving multi-word expressions, POS-aware lemmatization, and statistical collocation extraction—he…
Data poisoning attacks are a growing threat to machine learning pipelines, where adversaries subtly manipulate training data to compromise model behavior. Attackers use diverse techniques to inject malicious examples tha…
Indonesia plans to integrate artificial intelligence into its largest state initiatives, starting with a $15bn free-meal program for 83 million children and pregnant women, according to The Next Web. The government's nea…
A Hacker News user asked for AI-native alternatives to the spaced repetition app Anki, seeking tools that integrate artificial intelligence for improved learning.
Researchers trained deep MLPs to memorize secret binary strings and found that naturally learned behaviors resist detection by both human-made input-optimization algorithms and learned algorithms, posing a challenge for …
Engineers are increasingly using machine learning surrogate models to accelerate computational fluid dynamics (CFD) and finite element analysis (FEA) simulations, replacing traditional hour-long solver runs with faster p…
A benchmark inside an iOS app compared direct speech-to-tool processing against a two-step speech-to-text then text-to-tool pipeline for turning spoken commands into JSON tool calls. The direct approach avoids an interme…
A developer traces the 70-year history of AI from symbolic systems to modern transformer-based models, highlighting key milestones like the 1956 Dartmouth Workshop, expert systems, the first AI winter, the rise of statis…
Researchers introduced Reflective Masking (RM), a post-training method that enables Mask Diffusion Models to iteratively revise their own outputs by re-masking uncertain tokens and remembering previous attempts. Tested o…