Natural Language Processing news and analysis on Web Pulse: 5575 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
Apple's watchOS 27 developer beta introduces Siri AI, which unifies the assistant experience across iPhone and Apple Watch, enabling more complex wrist-based computing tasks like referencing notes and checking off remind…
Apple released the first public beta of iOS 27, which includes an opt-in beta of the revamped Siri AI that can perform complex tasks like looking up concert lineups and adding events to calendars from emails. The new Sir…
Meta's AI-powered Ray-Ban Meta and Oakley Meta smart glasses are transforming sightseeing with hands-free travel features including real-time landmark information, live translation in six languages, and first-person cont…
AI chatbots are redefining companionship, raising questions about their impact on human relationships. Critics argue digital surrogates may erode genuine connections, while proponents see them as a soothing presence for …
MarkTechPost published a tutorial on building a VideoAgent-style multi-agent system for video editing tasks, including intent parsing, graph planning, and tool routing. The system integrates FFmpeg, Whisper transcription…
A developer proposes a paradigm shift in deepfake defense, moving from reactive detection to proactive authenticity verification through a system that creates and validates cryptographic 'digital twins' of individuals ba…
A new essay on the history of language and computation argues that the power of words has always belonged to those who can name, order, and test them, from the ancient medical use of 'abracadabra' to modern AI prompting.…
A developer built PathFindr, an AI-powered campus navigation app using Expo, React Native, and Sarvam AI. The app combines interactive maps, intelligent search, and a conversational AI guide to help users navigate univer…
A toy model from Anthropic finds that summarisation can help humans oversee large volumes of automated alignment research, provided the agents are not scheming. In experiments with synthetic languages, stronger summarise…
Apple's new SpeechAnalyzer engine, introduced with iOS 26 and macOS 26, beats OpenAI's Whisper Small on clean English read speech with a 2.12% word error rate on LibriSpeech test-clean versus Whisper Small's 3.74%, accor…
Tonic AI released PrivacyBench, an open benchmark for de-identifying semi-structured data from work tools like email and Slack, which introduces novel metrics for evaluating synthesis quality. The benchmark uses 21 disti…
AI glasses from Meta, including Ray-Ban Meta and Oakley Meta models, can serve as a hands-free travel assistant for sightseeing, offering features such as landmark identification, translation, directions, calls, and mess…
Transept.ai shares field notes from localizing an AI translation product into 10 languages in 3 days, detailing the technical and logistical challenges of rapid multilingual deployment.
Samsung will launch AI smart glasses in autumn 2026, developed with Google and designed by Gentle Monster and Warby Parker, to compete with Meta's Ray-Ban glasses. The Android XR-powered glasses will feature dual front-f…
A large-scale evaluation of uncertainty estimation across 22 languages using two human-curated Q&A datasets and nine methods found that prompting models to reason in English, even for low-resource language questions, sig…
Waze is integrating Google's Gemini AI assistant into its navigation app to enable voice-based traffic reporting and destination search, aiming to reduce distracted driving. The conversational reporting feature, first re…
Large language models have advanced to the point where they can solve open mathematical problems, generate accurate QR codes, and handle real-time customer service calls, according to a compilation of examples from resea…
A new training-free approach called TCLA (Training-Free Class-Level Adaptation) enhances medical vision-language models' performance on out-of-distribution data by correcting inference logits with a few support samples, …
A new study reveals that the true energy burden of Vision-Language Models lies not in visual processing but in the generation of output tokens, with each output token consuming 11 to 39 times more wall-clock time than ea…
DeshwalX released electra-small-prompt-injection-v1, a 14M-parameter multi-label classifier based on google/electra-small-discriminator, achieving 90% Micro F1 and 89% Macro F1 on the WildGuard test benchmark for low-lat…