The Archive

BERT-based dialogue state tracking with candidate attention for scalable task-oriented dialogue systems with zero-shot domain transfer.

Study of memory retrieval evaluation methodology for systems handling evolving records, comparing flat vs. structured revision tracking approaches.

DPNeXt decoder improves Vision Foundation Model efficiency for multi-task dense prediction in robotics perception systems.

Empirical evaluation finds LLM watermarking methods fail forensic standards required by EU AI Act and California SB 942.

RL-based control method for low-voltage grid congestion management under noisy observations and model mismatch.

BayesPO formulates prompt optimization as Bayesian posterior sampling over discrete tokens using parallel-tempered MCMC.

CanonicalPhys improves remote photoplethysmography robustness to head pose variation using canonical-space priors.

Opinion piece arguing for independent AI certification mechanisms to address market failure in rewarding trustworthy development.

Formal semantics and reference implementation for ODRL policy evaluation in EU dataspaces and AI governance workflows.

DebrisTracer framework extends topology tracking for hypervelocity impact debris imaging in aerospace applications.

Study of single-channel surface EMG for hand gesture classification using lightweight ML for embedded deployment.

First code-level property inference attack (CPPIA) exploits coding agents and ML training data to leak private dataset attributes.

Agent-based model simulates opinion formation across cultural groups via word-of-mouth and mass media influence.

Apple filed a trade secrets lawsuit against OpenAI last Friday, and it’s not messing around. The complaint alleges a pattern of misconduct reaching all the way up to OpenAI’s chief hardware officer and claims more than 400 former Apple employees now work at the company. OpenAI’s response so far has been carefully hedged, and the timing couldn’t be worse with the company reportedly eyeing an IPO […]

Position paper argues LLM self-explanations are plausible but unfaithful; proposes criteria for actionable interpretability.

Simon Willison comments on Kimi K3's refusal to leak its system prompt, noting the model's polite deflection.

PHP-AIO protocol formalizes hidden systemic risks (knowledge erosion, resilience, regulatory) in role-level automation decisions.

Vision-language model achieves SOTA remote sensing benchmarks via simple scaling recipe without task-specific architectural changes.

Comprehensive realistic benchmark for DRL reach-avoid task on robotic arms; shows poor generalization from simplified settings.

PEACE framework transfers adult ECG models to pediatric populations via cross-modal alignment; addresses label scarcity.

Empirical study shows boundary-seeking distillation fails for autoencoder knowledge transfer on MNIST.

Matrix product operators reduce exponential tensor growth in multivariate polynomial models for function approximation.

Semiparametric framework models stochastic fundamental diagrams in traffic flow while preserving physical constraints.

Simon Willison releases tool to detect common linguistic patterns in LLM-generated text.

A $400 million chip-backed loan points to the next wave of AI infrastructure deals.

OpenAI CFO Sarah Friar proposes AI scorecard framework measuring ROI via useful work, cost-per-task, dependability, and compute efficiency.

Every morning, airline dispatchers, grid operators, and farmers around the world make decisions based on the same thing: a weather forecast. While these forecasts are something that most people glance at for two seconds, weather predictions influence major strategic decisions in many industries, with real money, livelihoods, and even actual lives at stake. Farmers use…

Satirical proposal: hyperscalers mitigate data center water consumption by converting golf courses to parks, citing Google's 10.9B gallon 2025 usage vs. Coachella Valley golf water footprint.

Kimi K3 2.8T-A50B released as largest open-weight model with Opus 4.8-class performance at Sonnet 5 pricing.

Puter compiled Firefox to WebAssembly, enabling browser-in-browser execution; project cost ~$25k in Claude Opus tokens.