Machine Learning News

UC San Diego researchers developed DFlash, a speculative decoding method that uses a lightweight block diffusion model to draft entire token blocks in parallel, achieving up to 6.08x speedup on Qwen3-8B and up to 15x thr…

A data science article explains the critical role of data distributions in statistical analysis and machine learning. It describes how distributions reveal the spread, center, shape, and outliers of data, and covers comm…

A developer released World Model MCP, a memory layer for AI coding agents that uses a temporal knowledge graph to prevent repeated mistakes, achieving a +10.2 point improvement on the SWE-bench Verified benchmark across …

A developer built an 11-chapter interactive guide explaining how LLM inference works, centered around nano-vLLM, a 1,200-line Python reimplementation of the vLLM serving engine. The guide covers algorithms like PagedAtte…

A developer demonstrates that using an LLM agent to pick A/B variants is flawed because the model lacks concepts of sample size and exploration, leading to suboptimal decisions. Instead, they show how to route the decisi…

Nexus Labs' fine-tuning and evaluation team lead demonstrated that a single evaluation metric like 84.2% accuracy on a 500-example set carries significant uncertainty, with a 95% bootstrap confidence interval spanning ro…

Insurance Journal reports that insurers are shifting focus from debating generative AI's role to identifying signals for scaling pilots. Executives at Carrier Management's InsurTech Summit said readiness is indicated by …

Photoroom reduced inference latency of its convolutional segmentation backbone by 22% on A100 GPUs by switching to PyTorch's channels-last memory format. The change required only four lines of code and did not affect acc…

A developer built a Pluribus-style poker AI from scratch, implementing Counterfactual Regret Minimization (CFR) and its variants to solve imperfect-information games. The project achieved Nash equilibrium convergence in …

A toy transformer retains provably predictively defunct belief state data in its residual stream, shedding it only under sufficient capacity pressure. Experiments extending Shai et al. 2024 show that information about a …

A developer built Signal-OSS, a GitHub Action that scores new issues from 0 to 10 using heuristics and an optional LLM tiebreaker, automatically posting tailored checklists to help contributors improve low-quality report…

Deltatensors, a new open-source tool, compresses fine-tuned neural network model deltas into small .wdelta files, achieving near-lossless compression with sub-1% perplexity difference. Tested on Qwen2.5-0.5B fine-tuned o…

An ML engineer built a system that automatically rolls back machine learning models before they harm production data by routing a small percentage of live traffic to a canary model, continuously measuring its error rate …

Agencies including WPP, Dentsu, and Butler/Till are racing to develop AI agents for media buying and planning, aiming to gain a competitive edge. However, brands like Hyundai are using the same technology to bring market…

Researchers developed a hybrid machine learning approach combining EfficientNetB0 feature extraction with Random Forest classification to achieve 94.58% accuracy and sub-millisecond inference for real-time melt pool moni…

Researchers propose ModTGCN, a modularity-aware graph neural network for text classification that jointly optimizes cross-entropy and a modularity-based auxiliary objective to promote class-coherent document communities.…

Researchers introduced QuechuaTok, a benchmark evaluating tokenization strategies for Southern Quechua, a low-resource agglutinative language. They found that BPE achieved the lowest fertility rate but only 6.67% morphol…

Researchers developed an automated pipeline to explore heterogeneous 4-expert mixture-of-experts architectures, generating 4,463 candidate models over 28 days. The search was biased toward the AirNet family due to an enu…

Researchers developed low-power analogue neural networks with trainable nonlinear connections, inspired by Kolmogorov-Arnold networks, achieving efficient continuous control tasks like robotic kinematics and photovoltaic…

Researchers reconstructed GRACE-like terrestrial water storage anomalies back to 1940 using a spatio-temporal graph neural network (MTGNN) applied to South America. The model achieved high correlations and low bias, outp…