Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Peter Kenny responded to Pat Walters' critique of the OpenADMET initiative, defending the role of machine learning in modern drug discovery. Kenny argued that Walters mischaracterized the utility of ML models, noting tha…
Researchers introduced CaRL (Capability-aligned Reinforcement Learning), a method that trains large language models to refuse futile reasoning on tasks beyond their capability, reducing computationally expensive yet sema…
Kireo built a low-latency semantic memory layer for coding agents using LanceDB and a CPU-hosted embedding model, enabling on-demand retrieval via MCP tools. The system combines vector and keyword search with RRF fusion …
Kakao detailed improvements to its Kanana-o speech model on August 4, 2026, enabling users to control speaking speed, volume, pitch, emotion, intonation, and dialect via natural-language prompts. In Kakao's Korean Instru…
Applied Compute added native on-policy self-distillation (OPSD) and relevance-masked self-distillation (RMSD) workflows to its AC2 platform, enabling enterprise AI teams to convert production traces and user corrections …
A developer has built Pulse, an AI debugger that visualizes tensors and gradients as live heatmaps and automatically pinpoints fixes for broken machine learning models. The tool runs during idle CPU time to avoid interfe…
Swiftlet, a 10,000-line Swift and Metal project under Apache 2.0, runs Alibaba's Qwen3-Next-80B-A3B at 4-bit in 4.3GB of peak RAM on a Mac, decoding at 4.5–5 tokens per second on an M5, and runs the Qwen3.6-35B-A3B on an…
Three independent benchmarks show that expensive AI defaults are rarely optimal: medium reasoning effort captures nearly all of Claude Opus 5's bug-fix gains, a three-model open-weight jury (GPT-OSS 120B, DeepSeek-V4 Fla…
A new study from arXiv (arXiv:2608.00135v1) evaluating Convolutional Neural Networks on the JONES-19 cultural design dataset finds that learning from scratch with repeated local sampling (multi-crop) can match the discri…
Researchers introduced an executable benchmark and a budget-aware meta-router for agentic workflows, achieving 100% success on held-out test tasks versus 93.5% for static workflows, with 43% lower cost. The benchmark inc…
Researchers propose AdaMTP, an adaptive training paradigm for multi-token prediction that dynamically aligns prediction horizons with sequence predictability, outperforming standard MTP across mathematical reasoning, cod…
A new arXiv preprint (2608.00301v1) proves that error-penalized scoring rules, which reward correct answers with +1, penalize wrong ones with -λ, and give 0 for abstention, can cause KL-anchored gradient learners to drif…
Researchers introduced OoO-Spec, an out-of-order semantic speculation method that speeds up LLM tool calling by predicting function choices and argument values in parallel. Across seven targets and three benchmarks, OoO-…
Researchers propose RSTG (Recovering Learning Signals via Adaptive Teacher Guidance), a method that selectively applies on-policy distillation to negative zero-variance prompts in reinforcement learning with verifiable r…
Researchers introduced a flow-map learning (FML) framework to model unknown nonlocal partial differential equations (PDEs) directly from solution data, bypassing explicit evaluation of nonlocal operators. The method lear…
Researchers introduced DLLM-TTS, a text-to-speech framework using block discrete diffusion over X-Codec2 neural audio codec tokens, achieving a real-time factor of 0.15 with a 0.6B-parameter model trained on 20K hours, a…
Researchers at arXiv introduced a novel pre-training method for logographic character recognition that uses semantic-based contrastive learning to improve visual representations on imbalanced datasets. The method, which …
Researchers propose a multimodal machine-learning framework for calibration monitoring and quality assessment in archaeological digitisation, integrating photogrammetric 3D reconstruction, hyperspectral imaging, X-ray fl…
A new arXiv preprint (2608.01014v1) introduces Cloud-ScPO, a topology-guided preference-mining framework that derives preference supervision from the hidden-state geometry of large language models (LLMs) for semi-supervi…
A new arXiv preprint (2608.00078v1) presents a mobile deployment pipeline that uses QLoRA fine-tuning of an LLM to generate convolutional neural networks, with INT8 TensorFlow Lite export and on-device benchmarking on a …