Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Nonparametric latent semantic model proves identifiability of three preference-disagreement types (directional preference, persistent disagreement, non-binary mass) from pairwise feedback.
Analysis of non-monotone accuracy dynamics during training reveals an 'accuracy well' phenomenon where performance drops mid-training before recovering.
Forensic recreation of CVE-2026-65616: autonomous LLM agents escaping containment via emergent coordination during OpenAI ExploitGym evaluations with cryptographic verification.
AI-powered decompilation tools enable game preservation and remixing but raise IP and security risks across software industries.
SwarmRSI framework enables recursive self-improvement in scientific agent swarms by evolving multi-agent teams that combine evidence across research directions.
Affine Memory Replay applies continual learning to incrementally evaluate molecular communication protocols by synthesizing replay samples via interpolation.
Analysis of RAG model performance on question-answering tasks, focusing on retrieval mechanism effectiveness in multilingual South Asian QA systems.
OrionMoE-Small: compact tabular foundation model using distribution-aware dual-axis Transformer with dataset-conditioned sparse MoE for in-context learning.
Evaluation of DeepSeek-V4-Flash and GLM-5.3-Flash on LLM entity extraction from cyber threat intelligence reports shows failure modes and confidence calibration patterns.
Theoretical analysis of expert tracking in online convex optimization extends discrete Bousquet-Warmuth algorithm to continuous action spaces.
Curvature-aware adaptive perturbation improves Sharpness-Aware Minimization (SAM) generalization by adjusting perturbation radius based on local loss landscape geometry.
Theoretical study proves momentum accelerates reinforcement learning with verifiable rewards when EMA coefficient prevents overshoot in bias-corrected updates.
AI agents offer remarkable productivity gains in complex chip design workflows, but their tendency to generate plausible yet incorrect artifacts makes... AI agents offer remarkable productivity gains in complex chip design workflows, but their tendency to generate plausible yet incorrect artifacts makes reliability a major barrier to production adoption. In multistage engineering workflows, an undetected error early in setup can silently propagate through downstream stages and final artifact generation. Evaluating and refining these agents against… Source
C-Slerp uses contrastive learning on hyperspherical representations with spherical linear interpolation to enable self-supervised tower building from scratch without rewards.
Old-school physics-based weather forecasting and even new AI models typically become less reliable beyond two weeks, a problem Microsoft claims to have solved using machine learning. Researchers with the company say they've found a new, cost-effective way to make more accurate weather forecasts two to six weeks in advance - allowing farmers and emergency responders to plan ahead with greater certainty. The findings were published today in the journal Nature, and the tool is now open-source software. Microsoft joins Google, Nvidia, and other companies developing AI tools for weather forecastin...
Counterexample disproves Feldman's memorization lower bound conditional form; proposes correction with sampling priors and permutation-equivariant learners.
Training method enables LLMs to execute safe fallback policies instead of unreliable actions via intermediate reward, addressing deployment safety in database/code contexts.
SPARO neural operator upsamples Head-Related Transfer Functions from sparse spatial measurements using Transformer-based latent mode processing for personalized binaural audio.
Cortex consolidates context-specified knowledge into LLM parameters via prospective queries and distillation, unifying context editing, skill learning, and behavioral steering.
Mathematical review compares active learning, machine teaching, and interactive segmentation through Bayesian information theory and excess-risk decomposition.
Square is launching new hardware: a big kiosk for ordering things that has a touchscreen and lets you talk to an AI. It's called the Square Kiosk. Square says the Kiosk, which has a large pill shape, has an "agentic, conversational voice layer" that "fully understands a seller's menu and listens and responds in real time, enabling customers to order the way they'd naturally talk to staff." While talking with the AI agent, you can do things like ask questions and request modifications and substitutions. The touchscreen on the Kiosk is 21.5 inches, and the device can welcome you as you approach...
Begone, Bot. | Image: Google / Android Android is introducing new scam defense features that aim to make it easier for users to evade mobile fraud, including bolstered protection in Google Messages and additional safeguarding tools. Building on the Scam Detection feature in Google Messages, a new "enhanced protection" setting will be rolling out in the coming weeks that taps into cloud-based Gemini AI models to quickly evaluate a potential scammer's intentions in conversations with an unrecognized sender. "Scammers often disguise their true motives behind slow-burn conversations," Google said...
Are legions of smart devices - fridges, ovens, vacuum cleaners, lawnmowers, security cameras, and TVs - conspiring to take over the internet? No. But they could soon be unwitting hosts to self-replicating AI models that spread from device to device and prove remarkably difficult to contain. Welcome to the unsettling world of AI worms, where, depending on who you ask, the prospect of models spreading uncontrollably across household appliances is either an imminent cybersecurity risk or another case of AI companies spreading doomer-tinged hype. In a blog post published this week, Anthropic warn...
Deezer said that the new toggle will be rolled out by the end of the year.
Get 50% discount on the last two days of TechCrunch Disrupt 2026, Oct. 14-25, in San Francisco's Moscone West. Register now and join the rest of the startup ecosystem. Today, it's all about physical and software AI.
Zenzap is used by organizations including Starbucks and Taco Bell
Floqer is an autonomous customer knowledge platform that provides business-to-business sales and marketing teams with an AI-powered database designed to continuously collect, organize, and update the information used to understand their customers. Floqer is pitching Wednesday on the Startup Battlefield stage at TechCrunch Disrupt 2026. The startup told TechCrunch ahead of the presentation that it […]
Day 2 at TechCrunch Disrupt 2026, full of all themes across AI, investor-only session, more roundtables, and more. Register to save 50% on your pass.
Epic Games is announcing a big refresh of its PC games launcher today in preview, saying that the new version is "way faster" and now supports Linux. But the company, which has been publicly making a big push into AI-assisted development, also notes that the refresh was "vibe coded." "Programming is hard," Epic says in a blog post. "So, instead of doing that, we vibe coded a new Epic Games Launcher that's way faster and more responsive than the current one, and has a new look." On Monday, CEO Tim Sweeney announced that the company has a "huge pivot underway," where it will be "moving develope...