Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A new open-source framework called Persona Loop uses observed behavioral data to build audience personas that make falsifiable predictions, score them against real outcomes, and self-correct biases through a feedback loo…
A developer built VoltMem, a memory layer for LLM agents that assigns different protection weights to facts based on their volatility, preventing stale knowledge from persisting while allowing volatile facts to update ea…
A developer at Dremio details the renaissance in columnar file formats driven by AI workloads and hardware evolution. The analysis covers how Parquet's assumptions are breaking under modern demands, and dissects new form…
PyTorch 2.13 shipped July 8 with FlexAttention gaining native Metal support on Apple Silicon, delivering up to 12x faster performance than SDPA on sparse patterns, and a new fused loss function that cuts peak GPU memory …
Researchers at an undisclosed institution have developed a spectral filtering algorithm that learns to predict the behavior of marginally stable unknown nonlinear dynamical systems with vanishing error, as described in a…
Researchers at the University of Maryland found that the community size of early adopters and emoji semantics are positively correlated with the popularity of newly released emojis on Twitter. Analyzing a large dataset o…
Researchers from Singapore developed a Signed Belief Graph Neural Network (SiBeGNN) for topic modeling in conspiratorial discourse, achieving a cDBI of 8.38 versus 13.60–67.27 for standard methods. The team analyzed 553,…
A developer released Zer0Fit, an open-source project that wraps Google's TabFM and TimesFM foundation models as a Model Context Protocol server, enabling LLM agents to invoke zero-shot ML tasks with no training required.…
Researchers Duy Dang Phu and Thìn Đặng Văn have developed an NLP framework combining relevance filtering, topic modeling, and LLM-assisted interpretation to analyze corporate strategic behavior in the Opioid Industry Doc…
AMD has released a Triton-based optimization for video sparse attention on its ROCm platform, targeting Diffusion Transformers (DiTs) used in video generation. The implementation reduces the quadratic cost of full self-a…
AMD's open-source GEAK agent-driven framework automated the optimization of the DeepSeekV4 MLA kernel, achieving a 2.10x improvement in end-to-end throughput and a 3.71x reduction in time-to-first-token on an MI355 GPU w…
A developer has released RNAValidate, an open-source validator that checks AI-predicted 3D RNA structures against experimental data such as FRET, cryo-EM, and SHAPE/DMS probing. The tool applies five rules to flag struct…
A developer built a benchmark to test whether a local-first knowledge tool called IWE could serve as agent memory by storing memories as linked markdown files and retrieving them with graph-aware tools. The benchmark rev…
Avriz, a cloud platform for coding agents, has built a contextual bandit system that learns to route LLM calls to the cheapest capable model without risking user experience. The system trains exclusively on billing data,…
A technical article explains Direct Preference Optimization (DPO) as a simplified alternative to Reinforcement Learning from Human Feedback (RLHF) for fine-tuning large language models, collapsing the multi-model RLHF pi…
A developer explains how metadata filtering and reranking improve retrieval in RAG systems. Metadata filtering narrows the search space by using chunk attributes like chapter names, while cross-encoder models rerank resu…
Google Research introduced TabFM, a foundation model that treats tabular data predictions as an in-context learning problem, enabling predictions via a simple API call without traditional model training. The model reduce…
AI Apps API launched Adaptive Recall, a persistent memory system for AI assistants that uses cognitive science and machine learning to improve retrieval quality over time. The API offers features like adaptive retrieval,…
A developer released the first open-source training kernels for MiniMax Sparse Attention (MSA) on Hopper and Blackwell GPUs, enabling efficient million-token training with sparse attention. The kernels, built in CuTeDSL …
Reinforcement learning researchers are debating the 'one-step trap,' where models prioritize immediate rewards over long-term goals, potentially undermining the development of sustainable decision-making in AI systems. T…