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.
A new study finds that transformer language models implement a unified mechanism for handling multiple mental spaces—counterfactual, belief, fictional, temporal, and depictive—using a shared router over a value slot, con…
A study on training reasoning language models (RLMs) in Japanese, led by researchers using Qwen-3-Swallow-8B, finds that reasoning-language control is feasible but achieves performance at best on par with English-reasoni…
A new study from researchers at National Taiwan University introduces LocalSubs, an on-device English-to-Traditional-Chinese subtitle translation model optimized for low latency and privacy. The model achieves a 59.2% ti…
A study using GPT-2 models fine-tuned on impossible languages and then reconstructed to natural English found that recovered structures exhibit shorter dependency lengths than the original text, revealing an architectura…
A new study from arXiv finds that commercial search APIs for tool-using agents achieve similar answer accuracy (25, 25, and 26 out of 100 questions for Brave, Tavily, and Firecrawl, respectively) but differ sharply in ev…
Researchers from the Singaporean Home Team adapted an open-source spoken language model (SLM) to handle five speech tasks in Singapore's four official languages, creating HT-Moonstone (5B). The model matches or outperfor…
Large language models (LLMs) used to summarize clinical trial results frequently generate unsupported claims, posing risks for healthcare providers, patients, and payers, according to a new benchmark study. Researchers e…
Researchers at EPFL introduced Prompting-MammAlps, the first camera-trap benchmark for text-to-video retrieval, and a fine-grained method that achieved a 34% F1-score on 135 ecological queries with 775 candidate videos, …
Bilibili released Index-1.9B, a series of open small language models including four variants: Base, Pure, Chat, and Character. The 1.9-billion-parameter Base model, pre-trained on 2.8 trillion tokens, achieves an average…
Researchers have developed PolyInterview, an LLM-based platform for immersive mock interview practice that generates tailored questions from a job description and CV, conducts multi-turn spoken interviews with a lip-sync…
Researchers propose PTEI, a framework integrating MBTI and OCEAN personality traits into Large Language Models to enhance Emotional Intelligence, achieving up to a 4 percent accuracy increase when combined with Chain-of-…
A new language-model forecasting system for merger arbitrage, developed by researchers at arXiv, achieves the best performance on predicting outcomes of announced M&A deals, reducing class-balanced Brier score to 0.151 o…
A technical walkthrough demonstrates the arithmetic behind attention in GPT-2 by computing one token's contextual embedding from raw scores, showing that attention weights and values are produced by separate model compon…
DoorDash engineers built a system using LLM juries and multimodal AI to generate and optimize food metadata, improving the accuracy and relevance of dish descriptions and attributes across its platform. The approach comb…
AI agents can automate content marketing from research to publishing, reducing manual work from 6-12 hours per post to minimal human intervention, according to a guide on building agent workflows using Claude Code and au…
G-Eval, a prompting framework from Microsoft researchers that uses a large language model as a reference-free evaluator of generated text, scores open-ended outputs in a way that aligns with human judgment. The 2023 pape…
WANDR (Wide ANd Deep Research) evaluates research agents on 500 realistic data-collection tasks requiring broad entity discovery and evidence-backed claims, using a reference-free grader that re-fetches cited pages to re…
A prompt injection classifier trained on a 14M-parameter Electra-small discriminator achieves 0.990 F1 and 0.999 ROC-AUC on English WildGuardTest prompt-adversarial classification, but shows edge cases in non-English lan…
Apple released the iOS 27 public beta today, bringing Siri AI, performance improvements, and new Apple Intelligence features. The update includes a revamped Siri with back-and-forth conversation capabilities and private …
Waze announced a batch of updates this week, including a 'less chatty' mode, personalized navigation, and Gemini-powered voice search and reporting, all underpinned by Google's Gemini AI. The updates roll out Monday in t…