Natural Language Processing news and analysis on Web Pulse: 5440 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
A developer has published a beginner-level explainer on large language models, covering core concepts such as weights, parameters, transformer architecture, tokenization, context windows, and sampling controls like tempe…
A developer built a controlled experiment testing whether explicitly stated relational context between documents improves retrieval-augmented generation reasoning when retrieval is held constant. Using a synthetic corpus…
A developer's technical lesson explains that large language models are stateless prediction machines with no built-in memory, and that the appearance of conversational recall comes from the messages array, which resends …
Deepgram posted a Director of Text-to-Speech Synthesis Research role paying $213,000 to $328,000 per year, remote in the USA, to own the company's TTS research strategy, model roadmap, and production speech-generation mo…
Wispr Advanced Interfaces Lab introduced Canto, a real-time dictation speech model that achieved the lowest word error rate among all models tested on a 10-hour evaluation set of English-language Wispr Flow dictations fr…
A developer at DenixLabs deployed an AI voice assistant and chatbot across three dental clinics in Sofia, handling over 1,000 real patient calls and booking roughly 20 to 30 additional hours per clinic each month, mostly…
Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking in its API and Google AI Studio, adding native speech-to-speech capability for voice-first agents. Gemini 3.8 Live Extended Thinking currently ranks #…
A developer fine-tuned GLiNER large v2.5 (459M) to 0.83 F1 against Gemini 3.1 Pro's labels for named-entity recognition on Reddit knife comments, after zero-shot GLiNER scored about 0.65 F1. Gemini labeled 4,290 comments…
Fujitsu launched Aurora, an AI voice agent within its Digital Workplace Services Agentic Service Desk, which autonomously resolves IT issues through natural voice interactions in more than 50 languages. Aurora combines A…
A tutorial published on Scikit-LLM shows developers how to build a multilingual text classification pipeline using multilingual LLM embeddings and scikit-learn without training a separate model per language. The walkthro…
A paper posted to arXiv on 7 Feb 2026 by Francesco Cagnetta claims the first theory that quantitatively predicts data-limited neural scaling exponents for modern LLMs trained on natural language. The theory isolates two …
A developer published a technical explainer tracing the evolution of attention mechanisms in transformer models, from the original Multi-Head Attention (MHA) introduced in the 'Attention Is All You Need' paper to newer v…
A developer behind the Codenames AI web game added Simplified Chinese support by using an LLM to generate candidate translations while building deterministic validators and human review checkpoints around them. The devel…
A method called SpectralShift extends the effective context window of Gated DeltaNet, a linear attention architecture used in place of softmax attention for long-context modeling, by reparameterizing the model's spectral…
Google launched the Gemini 3.8 Live and Gemini 3.8 Live Thinking AI models, adding real-time audio transcription, reasoning and translation plus multi-step task handling, the company said on 17 September 2026. Google rep…
Researchers introduced REPAIR, a self-evolving data augmentation framework that iteratively synthesizes training data to resolve long-tail concept confusion in scientific dense retrievers, according to an arXiv paper (ar…
A confidence-gated hybrid that escalates only the least-confident predictions from a stacked ensemble to an LLM Pareto-dominates both pure systems on all three emotion-recognition datasets tested, reaching 0.620, 0.643 a…
A study of on-device tool routing over 600 Korean and English requests and a catalog of 70 local actions found that abstention, not tool selection, is where a neural component is required, according to the arXiv paper 26…
A new arXiv paper (2609.17804v1) reports that large language models solve grade-school math word problems through a four-stage sequential pipeline — Schema Abstraction, Operation Planning, Operand Binding, and Computatio…
A systematic benchmark of open-source instruction-tuned LLMs for key-value pair extraction found that models including Gemma, Mistral, Qwen2.5, LLaMA 3, and DeepSeek approach supervised layout-aware systems on clean text…