The Archive

MyoFlow: discriminative flow-matching for HD-sEMG gesture recognition with cross-session and cross-subject generalization.

Online scheduling framework for edge LLM inference balancing latency and load across heterogeneous servers for agentic AI.

The new SimpliSafe Video Doorbell Series 2 adds 2K resolution and dual band Wi-Fi. | Image: Simplisafe DIY home security company SimpliSafe is bringing its AI-powered proactive security feature to the front door. The new SimpliSafe Video Doorbell Series 2 launches today for $199.99 and works with the company's Active Guard Outdoor Protection (starting at $49.99 a month). This combines AI analysis with live agents to detect potential threats and respond proactively. When the camera detects suspicious activity - using a combination of on-device AI, cloud-based computer vision, and facial recogn...

Google releases interactive ATLAS dashboard for AI & Economy research data; open-access visualization of global economic-AI trends.

Their startup, Artificial Intelligence Underwriting Company (AIUC) has raised $40 million in a Series A lound led by Ribbit Capital, with participation from First Harmonic.

AI agent workflow using GPT-6 Astra optimizes 2D CFET thermal design, reducing peak temperature 1.67 K via automated geometry and heat-removal proposals.

Study of counterfactual self-explanations across LLaMA-3 and Qwen-2.5 shows model scale predicts explanation faithfulness in sentiment analysis and NLI tasks.

Intrinsic Robot Rewarding reuses VLA frozen encoders and demonstration endpoints to autonomously evaluate and improve robot policy outcomes.

Tool-assisted workflow identifies common student mistakes in mathematical modeling (formulas, equations, regex) via LLM analysis for targeted feedback.

KV cache reuse across standard LoRA adapters on Qwen3-1.7B backbone preserves quality while reducing serving cost for multi-specialist deployments.

Symbolic Separation grounds LLM agents in ontology-constrained Virtual Knowledge Graphs to reduce hallucination in multi-step database and operational queries.

Non-intrusive digital twin data models via randomized dynamic mode decomposition and deep learning reduce CPU cost for fluid dynamics simulation.

Semi-supervised XGBoost pipeline constructs multi-stage hepatocellular carcinoma genomic dataset for automated liver cancer classification.

Scaling-score conformal prediction extends distribution-free coverage guarantees to multi-target regression without specialized models beyond point predictors.

Two-parameter Riemannian metric family unifies Euclidean, Bures-Wasserstein, and affine-invariant metrics for covariance matrix optimization.

ICML: Multi-agent framework enabling adaptive, learnable memory policies for long-term agent conversations instead of static heuristic archiving.

EviScope: Counterfactual benchmark (960 tests) evaluating whether LLMs ground answers in correct evidence vs. producing correct outputs from wrong sources.

SAMPLESELECT: Few-shot audio classifier mitigating spurious foreground-background correlations via per-sample dynamic feature masking.

First systematic OOD evaluation of congenital heart disease segmentation across scanner/protocol/institution/modality shifts using ImageCHD.

Bio-inspired meta-learning: 13 evolutionary strategies for discovering which activation functions to assign nodes in indirectly-encoded networks.

Design space for AI-assisted work: 12-experiment comparison (N=917) of metacognitive monitoring interventions across timing, level, and cue source.

Hierarchical Planning Decoder: Neuro-symbolic goal inference from partial multimodal episodes for human behavior anticipation in assistive systems.

Topological signatures via persistent homology improve GNN expressiveness beyond Weisfeiler-Lehman 1-WL test for non-isomorphic graph distinction.

Audio-visual turn-taking prediction degrades significantly (up to 50%) on cocktail-party testbed vs. clean benchmarks; multimodal domain adaptation needed.

Ars previewed Mozilla’s report on how cheap open models caught up on capability.

Salesforce Koa is built on Nvidia's open-weight Nemotron model and is trained to do sales, marketing and customer-support tasks.

Riemannian gradient descent and Gauss-Newton methods achieve optimal sample complexity for nonconvex matrix completion.

Analysis of retrieval vs. extraction failures in multi-hop QA systems across three benchmarks reveals distinct error modes beyond passage retrieval.

Novel architecture using adapter banks as low-rank perturbations of recurrent networks for learning motor primitives in biomechanical control.

MASA framework breaks self-referential adaptation loops in wild test-time adaptation by anchoring model uncertainty to frozen multimodal LLM semantic descriptions.