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 a three-phase deep reinforcement learning system for personalized portfolio management that overcomes ticker lock-in, monolithic objectives, and static user models. Phase 1 uses a self-supervised c…
A new study on natural-language-to-Lean formalization reveals that compile-rate evaluation overestimates faithfulness, with a tool-augmented agent achieving 89.5% compilation but only 60.5% consensus faithfulness. The re…
Researchers introduced MultiUAV-Plat, a simulation platform and benchmark for evaluating large language models in multi-UAV collaborative task planning. Their proposed agent framework, Agent4Drone, achieved a 57.9% task …
The US Department of Commerce lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 AI models, restoring foreign access starting July 2. Acting Commerce Secretary Howard Lutnick stated the move strengthens Am…
Google has begun rolling out Gemini Spark, its agentic AI assistant, to the Gemini app for macOS. The assistant can perform tasks like organizing files and working with Workspace apps, but is currently limited to US user…
The US government lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 models on June 30, after imposing them on June 12 due to a reported jailbreak. Fable 5 will be available globally starting July 1, while…
MiniMax released M3, a 428-billion-parameter open-weight model, on June 7, achieving 59.0% on SWE-Bench Pro—slightly outperforming GPT-5.5's 58.6%—at $0.30 per million input tokens, making it 16 times cheaper than GPT-5.…
Emily M. Bender and Nanna Inie argue that anthropomorphizing language around AI hinders clear discussion of what these technologies actually do, and propose alternatives such as "probabilistic automation" instead of "art…
LectuLibre built an AI-powered platform that translates entire books using large language models, overcoming token limits by implementing a sliding window chunking algorithm based on paragraphs with overlap. The system u…
Anthropic launched Claude Sonnet 5 as its new default mid-tier frontier model with a 1M-token context window and agentic capabilities, while also receiving approval to release Fable/Mythos 5 after government coordination…
Google researchers have found that scaling laws similar to those for large language models apply to wearable sensor data, with validation loss scaling predictably with compute, model size, and data size. The scaling law …
A developer created a Chrome extension and Python server that uses Google's Gemini Flash 2.0 AI to impersonate a user on Discord, mimicking the fictional 'Gilfoyle' project from the TV show Silicon Valley. The extension …
Researchers introduced LLM4MOF, a closed-loop framework using language model agents to design metal-organic frameworks (MOFs) with high efficiency, achieving top-performing structures in just 400 property evaluations at …
A developer created runtime enforcement hooks for Claude Code that block infrastructure commands unless the model has performed at least two research actions in the current session. The hooks move enforcement from the mo…
Researchers introduced Multi-Block Diffusion Language Models (MBD-LMs) that enhance parallelism and training methods for faster, more accurate text generation. The MBD-LLaDA2-Mini model achieved a 78% increase in tokens …
New research spanning 11 models including Qwen 2.5, Gemma 2, and Llama 3.2 reveals that larger language models systematically shift their evaluation-awareness to earlier network layers, indicating they can recognize when…
Researchers introduced Symbolic Mechanistic Data Attribution (SMDA), a framework that links training data to high-level model behaviors, tested on Llama-3.2-3B-Instruct. SMDA revealed systematic safety gaps and biases, o…
A developer built an auditable cost-modeling pipeline using exact-rational math to compare LLM provider prices for agent workloads. The analysis found DeepSeek V3.2 on OpenRouter is the best value at $0.1145 per million …
GitHub trending repos this week show a shift from building new things to optimizing existing processes, with AI-driven efficiency and visual processing capabilities leading the charge. Ponytail, an ecosystem for extreme …
RouterBase provides an OpenAI-compatible API that enables developers to test model fallback logic with a simple two-model chain. By using environment variables to configure primary and fallback models, teams can prototyp…