benchmark audits can silently fail in five ways
a study shows that perturbation-based construct-validity audits can produce misleading conclusions due to hidden pipeline failures, and proposes a due-diligence gate to catch them.
aitag machine learning
summaries filed under machine learning
graft adds per-word pronunciation control to text-to-speech
a new method lets text-to-speech models pronounce tricky words correctly by using short audio hints, without changing the speaker's voice.
aipytorch monarch comes to amd gpus for fault-tolerant training
pytorch monarch now supports amd instinct gpus with rocm, enabling single-controller distributed training that recovers from node failures without full restarts.
aihow photoroom built the data pipeline for prx
photoroom details its data strategy for the prx text-to-image model, covering dataset assembly, captioning, and storage formats.
aidata scientists shift to ai oversight roles
data scientists now spend more time managing ai systems than building models, with new roles in governance, prompt engineering, and agent supervision.
aihugging face ml intern: your first ml agent
a cli agent that automates machine learning tasks from plain english descriptions, handling coding, training, and model publishing.
ailerobot v0.6.0 closes the robot learning loop
new release adds world model policies, reward models, a deployment cli, six simulation benchmarks, and faster data loading.
aismall language models power next-gen agents
small language models are replacing large models for repetitive agent tasks, offering speed, cost savings, and on-device privacy.
aigetting started with the claude api in python
learn to set up the claude python sdk, make api calls, handle responses, use system prompts, and stream output.
aipytorch test infrastructure explained
a guide to understanding pytorch's dynamically generated tests, opinfos, and ci sharding for contributors.
aishallow neural nets as smooth variational problems
a new formulation replaces discrete neural network training with a globally well-posed variational problem over parameter densities, enabling direct solution via a linear system.
aimeasuring lag relevance in time series forecasts
new model-agnostic method uses shapley values and ghost variables to measure lag importance in univariate time series forecasting.
aigoogle introduces tabfm for zero-shot tabular predictions
tabfm is a foundation model that predicts on new tables without training, using in-context learning and synthetic data.
aion-device ai hackathon shows local execution wins
teams built real-time apps on snapdragon phones using executorch, proving local ai's value for privacy, speed, and offline use.
ai10 agentic ai frameworks to know in 2026
a rundown of ten agentic ai frameworks, from langgraph to llamaindex workflows, with notes on their strengths and best use cases.
aihumanity's last exam is a distraction
a look at the hle benchmark, why it was created, and the divided expert opinions on its value for evaluating ai systems.
aifiltered mixture-of-generators for synthetic survival data
fogs selects plausible synthetic samples from multiple tabular generators to improve downstream survival model training on small clinical datasets.
aideep multitask learning for mixed outcomes with shared sparsity
a new method uses deep neural networks and rank-based optimization to handle mixed outcome types in multitask learning with shared predictor selection.
aisemiscope: tuning gains in semi-supervised security classification
a study disentangles whether improvements in semi-supervised learning for security come from tuning the classifier alone or from joint optimization with the ssl pipeline.
aigrpo, dr. grpo, and dapo are three settings of one dial
three popular language model training methods all adjust the same number: the standard deviation of correctness marks across sampled answers.
ai