The Archive

CoBRA learns tool-use boundaries in LLMs via counterfactual margins, optimizing when to invoke tools by estimating instance-level marginal benefit.

Covariance-corrected Mahalanobis distance for few-shot out-of-domain intent detection in conversational agents.

LLM-labeled NER training for Persian customer chat anonymization, comparing DeepSeek-V3, Qwen3, and GPT-OSS labeling quality.

Action-class diagnostic framework decomposes multi-turn tool-calling failures into miscalibration vs. execution modes across open and closed models.

Terminology-driven framework generates large-scale diagram instruction data for improving VLM performance on scientific diagrams.

Embedded conditional independence tests detect whether LLM-generated text carries information beyond source, applied to German parliament speeches.

Dataset for modeling iterative problem-solving tracks solver performance, error persistence, and strategy shifts across repeated attempts.

Fal's H3 Max Live enables real-time video generation faster than playback speed, potentially enabling live streaming applications.

Gilbert + Tobin law firm implements governance framework for ChatGPT Enterprise and Codex deployment with CEO oversight and human accountability measures.

Apple says it has evidence that a former employee destroyed evidence of data theft after learning he was under investigation.

LatchBio benchmarks Grok 4.6 biosecurity safeguards; model outperforms frontier competitors on adversarial bio task detection and refusal.

Graham Dumpleton releases Wrapture, a Python library for function wrapping, tracing, and mocking in testing.

Kākāpō population recovery milestone: 325 juveniles added after record 2024 breeding season; species rebounded from 51 in 1995.

Versions of OpenAI's ChatGPT and SpaceXAI's Grok will join Google's Gemini on the Pentagon's central portal for AI tools.

As frustration over AI influencers has been growing, Instagram is limiting the reach of undisclosed AI profiles.

OpenClaw 2.0 improves UX but doesn't address underlying security issues in popular agent framework.

The seed round for the app that provides real-time legal and policy guidance to officers was led by SignalFire and Las Olas VC.

Lawsuit: Anthropic’s torrenting totally screwed songwriters as AI songs top charts.

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. By now you’ve probably heard about last month’s major AI security incident, in which OpenAI agents escaped their sandbox and hacked into the AI platform Hugging Face while trying to cheat on…

Context-Aware Interleaved Batching improves WhisperX speech transcription by maintaining historical context across batched audio segments via VAD boundaries.

ECHO-OFTRL algorithm achieves constant individual regret in N-player games without polylogarithmic horizon dependence via optimistic follow-the-regularized-leader.

SUN Programs unify model-based control and learned policies by compiling task semantics into MPC costs, RL rewards, and transition guards for long-horizon manipulation.

Sharp approximation rates derived for neural networks using low-dimensional latent parameter generators with affine mappings to network weights.

Four-stage forensic audit protocol for black-box identity verification of anonymously-released frontier models via API fingerprinting and configuration reconstruction.

Configurable semantic chunking framework for biomedical RAG replacing fixed-size chunks with entity-preserving windows and trigger-centered extraction.

OntoAligner-Ensemble voting framework reconciles heterogeneous ontology alignment methods including lexical, structural, KGE, and LLM-based approaches.

Template Model Builder (TMB) framework simplifies implementation of neural network mixed-effects models via automatic differentiation.

DIASENTINEL on-premise multi-agent system for type 2 diabetes risk screening integrates calibrated prediction, EHR extraction, and ADA guideline verification.

Complexity analysis of determining compatibility between pairs of succinctly-encoded conditional probability distributions for probabilistic graphical models.