geometry framework tests if ai learns or just memorizes
a new mathematical model uses fiber bundles to detect when machine learning systems discover causal laws instead of interpolating data.
topic
research
the granularity paradox in time-series forecasting
finer temporal disaggregation inflates in-sample fit but degrades out-of-sample accuracy due to recursive error compounding.
small routing changes cut city traffic and emissions
a real-world study shows that modifying under 2% of trips in navigation apps can measurably improve driving speeds and reduce fuel use across entire cities.
narrative world model helps ai write long fiction
a new memory system uses narratology to track story state for long-form fiction generation.
a hierarchy of policy learning problems
a new framework organizes policy learning tasks by data requirements, from detecting improvement to minimizing regret.
cora: per-slice orthogonal rotation for svd-based low-rank adaptation
cora introduces per-slice coherent orthogonal rotations to preserve pretrained geometry in svd-based fine-tuning, improving parameter efficiency.
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.
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.
automated data readiness for scientific ai
a new open-source framework called redi automates the transformation of raw scientific datasets into ai-ready formats with built-in reproducibility and scaling.
when pairwise comparisons fail for pluralistic preferences
local pairwise comparisons can misrepresent preferences when people hold multiple global priorities like fairness or proportionality.
small 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.
quantum mechanics powers lasers, secure comms, and future engines
a century after puzzling physicists, quantum mechanics now drives lasers, cryptography, and emerging quantum computers, with new research pushing into energy, biology, and gravity.
open source ai gap map released
current ai launches a detailed index of 421 open source ai products, with mit-licensed data on github.
shallow 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.
measuring 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.
agentic search for federated learning algorithms
auto-fl-research uses coding agents to automatically propose and test federated learning algorithm changes under fixed compute budgets.
wiola: a new small language model architecture from scratch
wiola introduces five novel components for efficient small language models, including spiral positional encoding and adaptive token merging.
path planning for air traffic control with human-friendly design
a new algorithm generates all safe aircraft paths to help controllers make faster, more transparent decisions.