Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers found that induction heads in transformers implement two smoothing mechanisms—soft context-matching and BOS token pseudo-counts—that interpolate n-gram estimates, enabling in-context learning to regularize ra…
Researchers revised the GestaltMML paper on arXiv, describing a Transformer-based multimodal model that combines frontal facial images, demographics, and clinical text to improve rare genetic disease diagnosis accuracy. …
Researchers applied federated learning to object detection for drones, enabling collaborative model training without centralizing sensitive aerial imagery. Using the Sherpa.ai platform and YOLO26 nano model on the KIIT-M…
Researchers introduced GRAFT, a per-word pronunciation conditioning mechanism for zero-shot text-to-speech that uses a short spoken sample to control pronunciation of specific words. In a blind English listening study, G…
Swipe Cleaner, an iOS app built on the OpenNomos ecosystem, processes all photo cleaning tasks on-device using Apple's Core ML and Vision frameworks, ensuring no user data leaves the device. The app employs a Tinder-styl…
SGLang development is being augmented with agent-assisted workflows that encode procedural engineering knowledge into executable skills, covering LLM serving, GPU kernels, diffusion pipelines, and production incident han…
Researchers introduced Decomposer, a post-training framework that decompiles symbolic music (MIDI) into editable Strudel programs. Using synthetic data and reinforcement learning, the system outperforms closed-source LLM…
A developer built Candidate Compliance Agent, a multilingual RAG system that lets users query Tamil Nadu election candidate affidavits in English or Tamil. The system extracts text from scanned PDFs using Tesseract OCR, …
Aureka, led by Dr. Weian Zhao, released Open Drug Discovery Engine (OpenDDE), an open-source biomolecular foundation model for AI-driven therapeutic discovery. The model focuses on co-folding and structural reasoning, bu…
A tutorial demonstrates building an autonomous AI co-scientist for EGFR C797S inhibitor discovery using ChEMBL, RDKit, SHAP, and BRICS. The system mines IC50 data, trains a scaffold-split Random Forest QSAR model, interp…
General Intuition and Kyutai released MIRA on July 6th, a playable multiplayer world model that simulates a four-player Rocket League-style match in real time using a 5B-parameter diffusion transformer, without a traditi…
Hugging Face released LeRobot v0.6.0 on July 7, expanding the open-source robotics toolkit with world-model policies, vision-language-action models, reward-model support, six simulation benchmarks, a rollout CLI with hum…
Google DeepMind's AlphaFold 3 predicts interactions among proteins, DNA, RNA, ligands, and other biomolecules, while Isomorphic Labs introduced the Isomorphic Drug Design Engine (IsoDDE) for drug discovery, signaling a s…
Google Cloud demonstrated elastic training on TPUs, where a worker failure during multi-node LLM training was recovered in under two minutes without restarting the job. Using the JAX AI stack (MaxText and Pathways) on GK…
Researchers introduce PuzzleMoE, a method for compressing large Mixture-of-Experts models via fine-grained element-wise merging and bit-packing, achieving up to 16.7% higher accuracy on MMLU at 50% compression and 1.80× …
Jackrong released an open-source knowledge base for LLM fine-tuning, dataset distillation, reinforcement learning, and local deployment. The guide provides reproducible training pipelines, SFT and RL workflows, data prep…
Roboflow released a tutorial demonstrating predictive maintenance with vision AI, combining RF-DETR and Gemini 2.5 Pro to detect bearing defects before failure. The system uses computer vision to continuously monitor equ…
Alibaba-NLP released the gte-reranker-modernbert-base, a ModernBERT-based reranker model under the Apache-2.0 license, designed for RAG and search reranking workflows. The 1.1 GB model is available on Hugging Face and ca…
Google's Vision Transformer base model, google/vit-base-patch16-224, is now available on Hugging Bay as a compact resilience fallback for high-demand AI artifacts. The 330 MB model, licensed under Apache 2.0, supports im…
BAAI released the bge-reranker-v2-m3 model under Apache-2.0 license, a 2.2 GB reranker for improving retrieval-augmented generation quality and production search stacks. The model has 500 upstream downloads and is availa…