The Archive

ThetreStageAI retrieval-augmented system for Bengali theatre production planning using multi-view embeddings and Gemini integration.

Nvidia’s new SoC is available in Windows laptops from six different manufacturers. We’ve collectively tested four of them so far. | Photo: Antonio G. Di Benedetto / The Verge Nvidia's RTX Spark N1X processor is finally here, and we've been collectively testing a small fortune's worth of powerful laptops sporting this new "superchip." RTX Spark is designed to be a high-performing Arm-based SoC for laptops and small desktop computers, rivaling the best from Apple, Intel, AMD, and Qualcomm. It has loads of cores, Nvidia RTX Blackwell graphics, and enough power and unified memory to run sizable l...

Actor and angel investor Mark Wahlberg says he's OK with using AI in filmmaking, provided it's used "to enhance what you're doing."

Meta's new terms appear to give it permission to crawl and scrape websites "associated" with the businesses that use its platforms - including online publishers, according to a report from Nieman Lab. The new terms, which go into effect on October 30th, say Meta can use this outside content to "develop, improve, research," and "train" its AI products. The new policy applies to business accounts that promote their content across Facebook, Instagram, WhatsApp, and Messenger. As pointed out by Nieman Lab, these terms seem to allow Meta to bypass a licensing agreement and train its AI models on c...

Watch an exclusive clip from Super Scam Me where a victim tracks down her scammer.

Counterfeits of luxury goods are a major problem for fashion houses. Counterfeiters are getting better at replicating brands’ marks and passing off dupes, known as “superfakes,” as genuine items. A startup called Veritas wants to tackle the problem with its own chip that verifies goods. Veritas was founded by Luci Holland, a former Tesla technical […]

Replit CEO Amjad Masad envisions a future of “self-driving companies,” where AI agents handle day-to-day tasks and people focus on higher-level decisions.

LLMs trained on experimental feedback learn to propose model improvements without expensive RL, enabling recursive self-improvement for AI research automation.

Theoretical analysis shows low-rank bottlenecks in frozen activations explain why LoRA parameter-efficient fine-tuning works without NTK linearization.

SchemaShield-Lite repairs LLM agent function-calling robustness against interface drift via schema validation and retry logic without model retraining.

SETAREH Transform sparsifies high-dimensional vector operations via Haar-wavelet sorting for reduced memory and communication in distributed deep learning.

Transfer learning via ridge regression with L2 regularization achieves minimax-optimal scaling laws on downstream tasks relative to source pretraining.

ML-based imputation reconstructs missing HiRISE RED4 spectral channel data for Mars orbital imagery after 2023 detector failure.

PyTorch Inductor operator scheduling misses predicted peak memory on 22/54 CPU and 19/50 GPU configs; profiling-based verification ensures reliable inference.

Ablation study isolates architecture vs. loss contributions in competing-risks survival analysis, decoupling confounded deep learning design choices.

AgentOnto guides LLM agents on enterprise analytics using ontology-structured knowledge and iterative feedback incorporation across databases and documents.

Untrained CNN features in wavelet space with coefficient-space regularization jointly reconstruct pressure and sound speed in photoacoustic tomography.

MemCompile decouples memory retrieval from evidence presentation for long-term agent memory, using query-conditioned compilation to reduce context redundancy.

SRCoT introduces a closed-loop Retrieve–Reflect–Revise pipeline for edge LLMs to generate evidence-grounded text with provenance tracking.

Token probability scoring method infers native language background from non-native English text using Llama 3.1 on three datasets.

MobileInject reveals cross-modal injection vulnerabilities in MLLM-based mobile agents through adversarial multimodal content embedding.

SimOPD transfers teacher knowledge in on-policy distillation via response similarity matching rather than token probability or hidden-state alignment.

Enterprise data assistant deployment study shows knowledge graph-based semantic mediation improves LLM data access for imperfect business user prompts.

Spatial-TRM uses recursive refinement and frozen VLM features to improve spatial context prediction and region grounding in natural images.

Optimist Deep Research models literature research as iterative state-centric search with mutable question decomposition and evidence weighting.

Active learning with conformal prediction maps transonic flutter boundary as elastic axis position varies, using Gaussian-process surrogates.

FlashReFocus separates depth recovery from blur rendering to enable interactive single-image refocusing with one diffusion pass.

Machine learning approach to optimize variational angles in quantum approximate optimization algorithms for combinatorial problems.

Theoretical analysis of ranking tradeoffs among classification metrics in binary classification tasks.

Neural control method to improve distributed networked systems during plug-and-play subsystem transitions while preserving stability.