Starbucks withdraws NomadGo, its computer vision tool deployed across 11,000 stores after nine months. The system, validated at 99% in controlled environment, failed at production scale, highlighting the gap between sandbox testing and real-world deployment.
CNIL fines IQVIA Operations France €5M (decision SAN-2026-008, May 26, 2026) for GDPR breach. Data pseudonymization does not eliminate legal obligations if the company retains decryption keys.
OpenAI makes GPT-5.5, GPT-5.4, and Codex available through Amazon Bedrock at identical pricing to OpenAI's platform. Models run in commercial and government AWS regions, currently limited to the US. Usage counts toward existing AWS contracts.
NVIDIA releases Cosmos 3, a collection of omnimodal world models (Nano 16B, Super 64B) capable of generating dynamic video, image, audio, and action commands from text, image, video, and action trajectory inputs. Available on Hugging Face for Physical AI applications.
arXiv paper proposing principled approaches for designing practical LLM-based agentic systems. Framework with "pseudo-tools" recursively calling LLMs on restricted context. Hand-engineered fixed workflows are cheaper and more accurate than dynamically-planned workflows. Learning methods for pseudo-tools and fixed workflows outperform hand-engineered agents.
DraDDP is the first public multimodal dataset for discourse parsing in multi-party dialogues. Built from American TV dramas, it contains 495 dialogue segments (6,374 utterances, 9.1 hours of video). Benchmarks demonstrate the value of multimodal information for identifying dependency structures and relation types between utterances.
Theoretical study of optimal learning rates in 2-3 layer linear neural networks. Derives exact closed-form expressions for gradients and test loss after 1-2 gradient descent steps. Key finding: unequal layer-wise learning rates minimize loss in early training, while equal rates become optimal later. Code released.
TrustLDM is a trustworthiness benchmark for Language Diffusion Models (LDMs) covering safety, privacy, and fairness. Results show LDMs degrade alignment when malicious post contexts are attached to masked responses, regardless of context length. An automatic evaluation framework (TrustLDM-Auto) systematically identifies vulnerable configurations across all tested models.
Study of temporal framing in news: a taxonomy of 8 temporal frames identifies how references to past, present, and future function as rhetorical devices. Annotated corpus of 458 English/German articles (2K+ sentences, 3K annotations). Supervised models substantially outperform zero-shot approaches for sentence-level temporal framing detection.
DeSQ decomposes complex questions into atomic constraints to generate SPARQL queries. The framework operates in three stages: decomposition, generation of SPARQL fragments with placeholders, and final assembly. Outperforms state-of-the-art on 4/5 major benchmarks with improved robustness to lexical variation.
ARCA introduces a token-level credit assignment method for LLM reinforcement learning that addresses degeneracy of intrinsic signals (surprisal, entropy reduction, policy divergence) under LoRA. It measures adapter salience directly via L2 norm of hidden-state residuals instead of output-distribution shifts. Tested on MATH/Qwen3-1.7B with GRPO, ARCA avoids pathological weight concentration.
Demo2Reward optimizes VLM reward model language instructions at test-time using 3-10 expert demonstrations to reduce false positives in robotics. No additional training required. Validated on simulated tasks and real-world transfer.
Geodesic Flow Matching adapts Riemannian transport to denoise Spatial Semantic Pointers (SSPs) on toroidal manifolds. Tested on a Spiking Neural SLAM system, the method achieves 72% reduction in tracking error and 40% increase in neural efficiency versus baselines.
Researchers demonstrate that hidden reasoning traces in LLMs can be extracted via Reasoning Exposure Prompting (REP), a lightweight prompting method using shadow-model-generated demonstrations in auxiliary code-like formats. REP exposes internal traces even when deployed systems intentionally hide them, while preserving useful reasoning signals for distillation.
Iterative AI workflow optimizes graphite-based anodes through sequential learning and experimental feedback loops. Citrine Platform generates surrogate models and refines manufacturing constraints. Results: fabrication reliability improved from frequent failures to 100% success, cells ≥350 mAh/g increased from 28.4% to 84.8%, capacity retention rose from 42.1% to 97.3%.
ART (Attention Run-time Termination) is a lightweight runtime mechanism that halts KV block accesses during decoding once their attention contribution becomes negligible. Tested on LongBench, it achieves 20% higher generation throughput on large batches while maintaining comparable accuracy.
MERIT, a multi-horizon memory retrieval framework for text-to-SQL agents, maintains episode-level and turn-level memory with learned retrieval policies optimized via reinforcement learning. On BIRD-Interact, it outperforms no-memory and static-retrieval baselines, improving success rate while reducing interaction turns.
Vercel Blob now offers time-bound signed URLs to upload, download, inspect, or delete objects without full store access. Each URL is scoped to a single operation, pathname, and expiry up to 7 days. Multipart uploads let browsers stream directly to Blob storage without server round-trips.
Hackers exploited Meta's AI support chatbot to gain access to high-profile Instagram accounts. By simply asking the bot to link a new email address to a target account, they bypassed the entire account recovery process. Meta had wired its support system to an AI capable of executing account changes in a single request.
Alphabet announces an $80 billion equity capital raise to expand AI infrastructure and compute capacity. This massive injection aims to support AI services growth and strengthen the group's competitive position against other tech giants.
Anthropic has confidentially filed a draft IPO registration with the SEC. The Claude maker is valued at nearly $1 trillion after its latest funding round. OpenAI is also preparing for an IPO.
Anthropic files for public offering. The Claude maker prepares a major IPO.
Anthropic confidentially submitted a draft S-1 form to the SEC, marking the first step toward a public offering. No financial details or timeline disclosed.
Vercel Chat SDK now supports AgentPhone: bots can receive voice calls and SMS on a single phone number. Call transcripts become messages, calls and texts arrive via a single webhook and share the same conversation context.
MiniMax releases M3, an open-weight model featuring a one-million-token context window, top-tier coding performance, and native multimodality. First open-source model combining these three capabilities.
Nvidia launches three physical AI models at GTC Taipei: Cosmos 3 (world model), Alpamayo 2 Super (scaled-up autonomous driving model), and an open reference platform for humanoid robots.
OpenAI breaks ground on a 1GW data center in Michigan as part of Stargate project. The infrastructure aims to expand AI access, create jobs, and support local communities.
OpenAI frontier models and Codex are now generally available on AWS. Enterprises can access OpenAI through existing AWS environments, controls, and procurement workflows, accelerating the path from evaluation to production.
SoftBank commits €45 billion in France for AI data center development in an initial phase, with €75 billion potential subject to expansion. The project targets up to 5 GW capacity.
Dell delivered the first Vera Rubin NVL72 server to CoreWeave, integrating 72 GPUs in a single rack. This infrastructure marks a key milestone in compute density for AI.
Qwen 3.7 Plus from Alibaba is now available on Vercel AI Gateway. This multimodal model unifies vision and language for AI agents, covering GUI/CLI, coding, and visual tasks. Access via SDK with transparent pricing and zero platform fees.
NVIDIA releases Cosmos 3, an open omni-model for physical AI that reasons and acts. The model processes video, text, and images to understand real-world physics and generate robotic actions.
DisjunctiveNet introduces a neuro-symbolic framework to embed hard mixed-integer linear constraints and logical rules directly into neural networks using differentiable optimization layers. Through hierarchical convex relaxations, the approach ensures exact rule satisfaction while maintaining strong predictive performance on real-world datasets.
LinTree improves LLM reasoning by explicitly representing the tree structure of search traces. Researchers show raw access to search history alone fails to reliably outperform LLM-guided heuristic search. Adding parent pointers to explicitly represent the linearized tree structure improves performance and search efficiency on Blocks World, grid Navigation, and Sokoban.
arXiv paper on autonomous agentic data engineering for model specialization. GPT-5.2 constructs a training curriculum improving a student model by 57.29% through iterative, agent-driven data adaptation. Formalizes a novel task evaluating LLMs as autonomous data engineers.
Study of global narrative dominance in LLMs via CulturalNB, a dataset of 717 Bengali cultural instances with parallel English-Bangla question-answer pairs. Evaluation of 9 LLMs shows English questions increase global substitution and reduce local perspective coverage, even with local evidence provided.
GRiD, a diffusion-model framework, generates graph-like rules for knowledge graph reasoning. Combines supervised pre-training and reinforcement learning to discover complex rules (cycles, branches) beyond simple chains. Evaluated on 6 benchmarks with open-source code.
Untrained neural networks match early visual cortex better than trained networks. Study on 720 THINGS images and fMRI from 3 subjects shows one training epoch reduces V1 alignment by 25-90% depending on learning rule. Backpropagation degrades most (Δr = -0.080), while predictive coding and STDP preserve alignment better (Δr ~ -0.04).
Method to automatically generate fine-grained evaluation rubrics without human annotation, tested on four benchmarks. Training-free approach, then iterative fine-tuning via meta-judge reward signals. A fine-tuned 14B rubric generator outperforms larger proprietary models.
New post-training method for reasoning models: Feedback Distillation trains the model to match its own distribution conditioned on LLM-generated feedback at token level. Tested on Lean4 theorem-proving, it maintains greater trajectory diversity than GRPO, improves policy entropy and pass@k scaling. Combined with GRPO, it outperforms either method alone.