The Archive

Sparse Landmark Embedding kernel eliminates conditional negative definiteness requirement for non-Hilbertian distance measures in GPs.

According to The Information, Apple is planning to get back into the server game and might just pair up with NVIDIA to make it happen. Apple retired its Xserve line in 2011 and has largely left enterprise machines to other manufacturers since. But the growing demand for compute power as the AI industry continues to expand has apparently led the company to believe there is an opening for its powerful but efficient ARM-based M processors. Its Mac Mini and Mac Studio have proven popular with AI developers, which has led to shortages. The server product, whether it's called Xserve or something el...

Unified framework for sparse, probabilistic linear explanations applicable to classification and regression via Boolean hypercube mapping.

Statistical mechanics explains double descent phenomenon in neural networks via Boltzmann distribution over energy landscapes.

RLLBC-Lib educational code library for tabular reinforcement learning and control, targeting student accessibility.

Incremental memory maintenance for language-model game NPCs avoids recomputing full prefix on memory updates.

Formalism for social laws in stochastic multi-agent environments with α-robustness verification.

VyPER uses hypergraph representation learning and graph-conditioned diffusion for particle collider event reconstruction.

Google is inviting third-party agents, including Claude and Open Claw, into Google Home. | Photo by Jennifer Pattison Tuohy / The Verge Google is opening up its smart home to AI agents, letting tools like Claude and Open Claw access and control your connected devices and analyze your home's data using the standardized Model Context Protocol. Google Home MCP is a new integration that lets third-party AI agents control and monitor your smart home and act on your behalf. It "allows any AI agents that support MCP, including Google Antigravity, Claude, Hermes or Open Claw, to securely work with al...

Google is launching early access to a new MCP server for Google Home, allowing AI agents like Claude, ChatGPT, and others to control connected devices, review camera summaries, and access smart home activity using natural language.

OpenAI publishes framework for tracking and disclosing model misalignment with six incident reports of unexpected behavior.

HOPE second-order pruning method for Mixture-of-Experts models reduces parameter count by modeling expert cooperation.

DualViewEval compresses agent benchmarks using six process signals beyond task outcomes.

ECtHR-NPD benchmark for predicting non-pecuniary damage awards from European Court of Human Rights cases.

Claim-level discourse framework for analyzing densely entangled health narratives in social media.

Physics-informed kernel methods achieve fast learning rates combining value and differential observations.

Anthropic is initially releasing these features to Pro and Max plan subscribers.

Claude is getting a pair of new tools today: Docs and Slides. They'll let you create documents and presentations through Claude chats, which you can export, edit, and share with other users. As part of the announcement, Anthropic is also simplifying how Claude chats work, merging regular chats and Cowork into "one Claude," with all of its AI productivity tools available from any chat. Artifacts and Claude Design capabilities will be available through the new single interface as well. According to Anthropic, Claude "can now figure out what a task needs, so what Cowork and Design can do is avai...

cuTile Rust (cutile-rs) is a tile-based system for safe, idiomatic GPU kernel authoring in the Rust programming language. Extending the Rust ownership model to... cuTile Rust () is a tile-based system for safe, idiomatic GPU kernel authoring in the Rust programming language. Extending the Rust ownership model to tile-based GPU kernels, it splits mutable outputs into disjoint pieces and preserves the host-side ownership contract across kernel launches. It also allows programmers to opt out locally when they need lower-level control… Source

Framework for learning Lyapunov operators that map vector fields to stability functions via PDEs, applicable to nonlinear dynamical systems.

Study quantifies minimum human-to-synthetic data ratio needed to prevent model collapse in LLM training using Fisher-Rao information geometry.

ASLEval framework measures privacy exposure in multi-step LLM agent sessions, distinguishing local evaluation proxies from true session-level disclosure.

Physics-informed model for crystallographic texture prediction in laser powder bed fusion of Inconel 718 under defocus conditions.

Mustafa Suleyman argues against attributing consciousness, preferences, or rights to AI models, warning that doing so complicates alignment and containment efforts.

PersonaPath system for knowledge-centric adaptive learning that reasons over learner profiles and prerequisite structures rather than exercise patterns.

Sparse autoencoders reveal Gemma-3 and Qwen3.5 contain encoded evidence for harmful meme detection but fail to route it to outputs, not lack evidence.

OpenAI and AARP launch free ChatGPT workshops for 1,000 older adults across 10 U.S. cities to teach practical AI skills.

Demonstrates multiple autonomous AI agents in O-RAN control plane exhibit unsafe emergent conflicts despite individually correct objectives; proposes stability-guaranteed arbitration.

GrainSpeech compact speech synthesis model reduces encoder context beyond 15 phonemes and applies gradient-variance supervision to improve acoustic prediction.

EviGen framework for verifiable clinical rationale from EHRs using patient-conditioned retrieval and evidence scaffolding to reduce LLM hallucination.