Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Researchers introduced LPES, a layer-specific positional embedding scaling method that mitigates the 'lost-in-the-middle' problem in LLMs by assigning distinct scaling factors to each layer, achieving up to 11.2% accurac…
Researchers at arXiv introduced a persona-driven rewriting pipeline that conditions user turns on low agreeableness to reduce jailbreak susceptibility and harmful output rates in LLMs fine-tuned for social warmth, preser…
Researchers propose a symbolic feedback-driven iterative self-refinement framework to improve the robustness and reliability of large language models in long-horizon planning tasks. The framework uses natural language pr…
Researchers propose HybridCodec, a novel approach combining discrete tokens with continuous residuals to improve speech representation in language models, reducing information loss and autoregressive steps while retainin…
Researchers found that personality prompting of LLMs affects multi-agent team performance only in open-ended and competitive tasks, not in structured coding. Low agreeableness degraded collaboration and bargaining outcom…
Researchers proposed DysLexLens, a low-resource LLM framework to analyze dyslexic learners' experiences with AI tools using Reddit forum data. The framework filters noisy posts, integrates knowledge-graph reasoning, and …
Researchers found that explicit reasoning does not always improve multimodal emotion recognition accuracy, with fast thinking often outperforming slow thinking. They propose MER-R1, a reinforcement learning framework tha…
Researchers propose Tree of Evidence (ToE), a hierarchical claim verification framework that uses reinforcement learning-driven multi-source retrieval and argument tree aggregation to combat fake news and AI-generated mi…
Researchers introduced NormAct, a benchmark evaluating whether multimodal large language models comply with hidden social norms during embodied planning. Tests on GPT-5.4, Claude Opus 4.7, and Gemini 3 Pro showed models …
JD.com launched Oxygen AIIC, an industrial-scale platform using LLMs/VLMs to manage item knowledge for tens of billions of SKUs. The platform achieves 94.2% precision and 82.8% recall in knowledge production, processes h…
Researchers propose Tandem Reinforcement Learning (TRL), extending the tandem training paradigm to reinforcement learning with verifiable rewards (RLVR). Training Qwen3-4B-Instruct on competition math, TRL matches vanill…
Carles Marin released an open-source bilingual guide on Transformer architectures, covering attention dynamics, KV-cache compression, and advanced concepts like grokking. The resource includes reproducible code and inter…
A developer describes how Model Context Protocol (MCP) servers have transformed their AI workflow, enabling AI models to interact with external systems like GitHub, filesystems, PostgreSQL, Slack, and browsers. The devel…
Researchers introduced PCB-Bench, the first comprehensive benchmark for evaluating large language models on printed circuit board placement and routing tasks. The benchmark includes over 3,700 questions across three task…
Researchers have developed a method for on-policy distillation (OPD) that works across different tokenizers, enabling knowledge transfer between large language models (LLMs) from different model families. The approach us…
A developer argues that the term 'AI agent' is overused and causing engineering mistakes, defining a true agent as a system with an objective that decides next steps, handles failure, and knows when it's done. The develo…
An open-source MCP server called codebase-memory-mcp, built by DeusData, cuts the token cost of structural code queries by up to 121x by using a persistent knowledge graph. The tool, trending #1 on GitHub, indexes codeba…
OpenAI restricted its new GPT-5.6 Sol AI model to customers approved by the Trump administration, citing government requests amid heightened national security scrutiny. The move follows similar actions against Anthropic,…
Llmaker, an open-source platform for self-hosting a complete LLM stack including models, vector databases, embeddings, caching, observability, and an agent layer, launched on Hacker News. The platform provisions and mana…
Tau is an educational Python coding agent designed to teach how agents are built by exposing every component—provider adapters, harness, coding environment, and terminal UI—in readable layers. The project emphasizes a cl…