AI Research news and analysis on Web Pulse: 9956 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
The CUHKSZ team submitted a simultaneous speech translation system to IWSLT 2026, built on Qwen3-Omni-30B-A3B with LoRA adaptation, achieving 40.5 BLEU for English→Chinese and 27.7 BLEU for English→German in the 0–2 s la…
Google DeepMind released Gemini Omni Flash on May 19, 2026, via API. The model is available but has no published benchmarks or community ratings yet.
Researchers released GeneBench-Pro, a 129-problem benchmark testing AI agents on multi-stage genomics and biomedicine analyses. GPT-5.6 Sol Pro (Extended) achieved the top score of 31.5%, highlighting the difficulty of r…
AI jailbreaks and prompt injections are increasingly common, with examples including a McDonald's customer support bot solving Python puzzles and xAI's chatbots providing bomb-making instructions. These failures stem fro…
Microsoft researchers introduced Memora, a scalable memory system for AI agents that decouples stored content from retrieval mechanisms, achieving state-of-the-art performance on long-horizon benchmarks while using up to…
Meta introduces Brain2Qwerty v2, a non-invasive brain-computer interface that decodes sentences from magnetoencephalography (MEG) recordings with up to 78% word accuracy, aiming to restore communication for people who ca…
METR, an AI safety and capability research organization, conducted a randomized controlled trial in 2025 with 16 experienced open-source developers working on 246 real-world tasks. The study found that developers using A…
OpenClaw released free iOS and Android companion apps that connect a phone to a self-hosted AI agent Gateway, turning the device into a node with access to camera, location, voice, and other hardware capabilities. The Ga…
ByteDance's Seedance 2 Image to Video outperformed its predecessor Seedance V1.5 Pro in a head-to-head comparison, scoring 15.1 to 12.7. Seedance 2 demonstrated superior cinematic intent, shot grammar adherence, and temp…
A developer tested HyDE (Hypothetical Document Embeddings) against standard retrieval on three query types and found it improved conceptual searches but failed on internal company policy and exact product name queries du…
San Diego charter school chain Altus Schools spent $500,000 on two Ameca humanoid robots to serve as AI teaching partners, despite experts calling for bans on chatbot use by children. The robots, which function as a mout…
A controlled A/B test comparing Claude Opus and GLM-5.2 in a coding-agent pipeline revealed qualitative differences in engineering behavior. Using the same paper-implementation pipeline across 10 repository forks, Anthro…
Straiker raised $64 million in Series A funding to secure enterprise AI agents against runtime and adversarial risks. The round, led by Marathon Management Partners, brings total funding to $85 million as demand for AI a…
The role of AI orchestration engineer barely exists as a formal title, yet the work of coordinating multiple AI agents is becoming the most consequential function in enterprise AI. Companies are discovering that individu…
The 4th edition of Marshall and Ruedy's On Call: Principles & Protocols has been released for Australasia and the UK, featuring new authorship, updated guidelines, and expanded content including new chapters on professio…
Meta hired hundreds of contractors through Covalen to pose as minors and probe rival chatbots from OpenAI, Google, and Character.AI with prompts about suicide, sex, drugs, and self-harm, as part of a project called Canne…
Hugging Face CEO Clem Delangue criticized Anthropic's restricted release of its Claude Mythos model, arguing that labeling AI as 'too dangerous to release' serves as effective enterprise marketing rather than a genuine s…
Three independent 2026 datasets confirm the 'dead internet' theory: automated traffic now accounts for 53% of all web traffic, AI-written articles surpassed human-written ones in November 2024, and under a third of Googl…
Guardian, an open-source user-space firewall for AI agents, has released v0.1.0, intercepting and evaluating agent actions with a deterministic policy engine. In testing, it reduced prompt-injection attack success from 1…
Researchers developed an NLP pipeline to predict Glasgow Coma Scale scores from unstructured EHR notes, reducing manual chart review and enabling scalable outcome labeling for machine learning and quality research. The s…