Natural Language Processing

Researchers have developed a system that automatically discovers an executable schema from raw multi-source data, using it as a shared contract for knowledge graph construction and query-time retrieval. The system improv…

Researchers have developed ReasoningFlow, a framework that maps the non-linear discourse structures of large reasoning model (LRM) traces into directed acyclic graphs (DAGs) to improve evaluation and monitoring. Through …

Researchers have developed PERSUASIONTRACE, a framework that tracks human belief changes across multiple turns in conversations with large language models, revealing that people cluster into two distinct groups of belief…

Researchers introduced generic triple-latent sequence models that maintain a running token state and compressed pair-memory pathway to capture higher-order token interactions without benchmark-specific parsing. The tripl…

Researchers have developed CVT-RL, a reinforcement learning algorithm that uses policy-conditioned counterfactual credit assignment to reduce unsupported evidence chains and shortcut actions in long-horizon language agen…

Researchers have introduced state commitment learning, a new training objective that teaches language models to distinguish temporary computation from persistent memory. The method, called Counterfactual Erasure RL (CERL…

Researchers introduced a two-stage early failure alerting system for dialogs and LLM-agent trajectories that learns from sparse turn-level evidence rather than assigning terminal labels to every prefix. The attention-bas…

Researchers have developed LANTERN, a lightweight memory layer that recovers facts lost when large language models compress long conversations, achieving 78.3% recovery of verifiable facts without requiring any LLM calls…

Researchers developed a variance-aware reward framework using Group Relative Policy Optimization (GRPO) to improve heart-focused medical question answering in large language models. The approach, which replaces tradition…

Researchers at arXiv propose a hybrid pre-training objective for language models that combines a Joint Embedding Predictive Architecture (JEPA) latent-space prediction loss with masked language modelling (MLM). Pre-train…

Boson AI released Higgs Audio v3 TTS, a text-to-speech model designed for voice chat that generates expressive conversational speech across 100+ languages with zero-shot voice cloning and inline control over emotion, sty…

Tokens are the fundamental unit of text that AI models process, breaking down words, symbols, and spacing into small pieces for reading and writing. The context window, which can now reach up to 1 million tokens in some …

A developer built a speech emotion recognition and stress prediction pipeline using Wav2Vec 2.0 and Transformer models, enabling AI to estimate cortisol levels and emotional states from vocal patterns. The system uses a …

A developer built an AI-powered web scraper that uses large language models to extract product data from e-commerce sites, replacing fragile CSS selectors and regex patterns. The approach converts raw HTML into a simplif…

A RAG pipeline that fails to model patient identity before retrieval cannot deliver clinically relevant answers, according to a system built at a Global AI Hackathon. HPPIE (Hyper-Personalized Patient Insights Engine), w…

A developer built a natural language search for an internal documentation site by calling an AI API instead of training a custom model. The approach uses a well-crafted prompt and a small set of relevant documents to map…

A daily user of large language model (LLM) chatbots reports that the AI consistently fails to complete complex tasks, delivering only "half-baked" answers that require significant human effort to finish. The user advocat…

Accounting firms are adopting AI meeting assistants to capture client advisory calls, transcribe technical tax discussions into plain-language summaries, and track action items across multiple engagements. The tools fall…

Large companies can reduce their AI costs by implementing a local LLM filter layer that routes simple queries to free, open-weight models before falling back to paid external providers like Claude or OpenAI. The approach…

Microsoft Copilot can generate meeting notes in Microsoft Teams without saving a recording, but the feature requires a paid Copilot add-on license and notes are deleted when the meeting window closes unless transcription…