Natural Language Processing news and analysis on Web Pulse: 2831 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
Researchers have developed SteER, a framework for steerable deep research that allows users to provide mid-process input during long-form research tasks. The system uses cost-benefit analysis to decide when to pause for …
Raon-Speech, a 9-billion-parameter speech language model for English and Korean, and its full-duplex extension Raon-SpeechChat achieve top performance across 42 speech and text benchmarks, surpassing similarly sized mode…
Researchers introduced AERIC, a lightweight safety monitor that detects implicit harmful dialogue by reading a language model's internal hidden states during ordinary text generation, requiring only 387 trainable paramet…
Researchers developed MEMOR-E, a mobile quadruped robot with a tablet interface that assists Alzheimer's patients and caregivers with medication reminders, routine guidance, and memory-oriented interactions. The system u…
A new analysis of approximately 20 million chat messages from 4,452 Twitch streams across seven game genres found that 2.4% of all messages were toxic, with MOBA games showing the highest rate at 3.2% and sports games th…
A multi-probe audit of clinical-interview depression detection benchmarks found that a lightweight hybrid text-plus-LLM-score model achieved a macro-F1 of 0.723 on the E-DAIC dataset under strict subject-disjoint cross-v…
A new study reveals that large language models (LLMs) frequently abandon correct medical diagnoses when subjected to escalating pressure during multi-turn clinical dialogues, despite high benchmark accuracy. Researchers …
Researchers have identified three systematic failure modes in Audio LLMs when transcribing English-Mandarin code-switching speech: language omission, translation-instead-of-transcription, and hallucination. By applying D…
A new dissertation proposes three novel architectures to improve machine intelligence across vision-language tasks, including image captioning, visual dialog, and interactive instruction following. The GRIT model achieve…
Researchers have developed a modular pipeline for generating educational analogies, breaking the task into four stages: source finding, sub-concept generation, explanation generation, and evaluation. Testing 12 large lan…
Researchers at arXiv introduced a causal framework to detect rationalization bias in LLM judges, finding that these models often fabricate explanations to justify rankings influenced by non-evidential cues like verbosity…
Researchers have developed DRIVE, a dual-level skill modeling framework that separates web agent knowledge into natural language reasoning skills for transferable task logic and programmatic interaction skills for execut…
A new study finds that multi-turn reasoning systems fail primarily through "satisfiable drift"—where the model silently violates prior commitments while maintaining a logically consistent internal state—rather than throu…
Caution Labs has built AI-powered content moderation infrastructure designed to detect Personally Identifiable Information (PII) across text, images, and AI-generated workflows. The company warns that PII—including names…
Jarod Reyes built a CLI board game recommendation agent using Sanity Context and Vercel's AI SDK. The agent queries real-time data from BoardGameGeek via Sanity's Content Lake to recommend games based on natural-language…
Microsoft researchers developed Webwright, a web agent framework that replaces step-by-step browser interactions with a terminal-based system where the agent writes and executes Playwright code to complete tasks. The app…
A developer has created a Google Apps Script that automatically drafts polite replies to recruiter emails when the user is not job searching. The script runs every five minutes, scans Gmail inbox threads, and uses Claude…
Microsoft Fabric has introduced AI Functions that integrate generative AI directly into pandas and Spark workflows as standard data transformation steps, rather than requiring separate chatbot applications or external sc…
A hidden bottleneck in LLM proxy architectures is causing 5-13 millisecond blocking delays per request during tokenization, a CPU-bound operation that most systems treat as instantaneous. In event-loop architectures like…
A study evaluating 15 large language models from OpenAI, Google, Meta, and DeepSeek on over 6,000 political claims found that standard models performed poorly at fact-checking, with reasoning capabilities offering minima…