LLM-Typesafe 0.1a0 is a new plugin for LLM that adds support for TypeSafe AI's Jev model, enabling yes/no and choice questions with scoring.

LLM-Anthropic released version 0.29, which adds support for Claude Opus 5.5. This update provides users with the ability to interact with the Claude Opus 5.5, a large language model.

OpenAI and Anthropic have released new models, GPT-6 Sol, GPT-6 Luna, and Claude Opus 5.5, with significant price reductions. GPT-6 Luna is one of the cheapest models OpenAI has released, and Opus 5.5 addresses user feedback on its communication style and is cheaper per token than Opus 5.0.

ai·rete·rag is a tool that combines a Rete rule engine with a language model to provide auditable decisions. It runs the rule engine first, then uses the LLM to generate a plain-English explanation of the decision. The tool has a visual editor and supports forward chaining and nested rules.

OpenAI released GPT-6 Sol and Luna, two AI models that bring frontier intelligence to everyday work with different capability and cost balances.

A research paper proposes G-RELIC, a method for relexicalizing clinical documents using a graph-based approach. The method combines LLMs with graphs to optimize entity correspondence and preserve temporal consistency, outperforming state-of-the-art baselines in relational integrity and temporal coherence.

TypeSafe AI introduced Jev, a new type of Large Language Model (LLM) that returns floating-point numbers for categories, yes/no questions, ratings, and confidence scores. It's cheaper than OpenAI's GPT-5 Nano and can be used for classification tasks, search reranking, and other applications. However, concerns about bias and lack of transparency are raised, emphasizing the importance of evals and structured experiments.

A new plugin, llm-keys-ui, has been released. It allows users to securely manage API keys for LLM projects without pasting them directly into the ChatGPT app. The plugin uses Codex to run commands and retrieve key URLs for local network or Tailscale device IPs.

A dataset of teacher-written case studies for evaluating LLMs' ability to predict teacher interventions in classroom settings.

Researchers propose SCGFM-ART, a structure-centric framework for graph foundation models that aligns arbitrary graphs onto a shared relational atlas, enabling transferable representations across heterogeneous domains. This framework achieves state-of-the-art transferability and significantly speeds up inference time by avoiding iterative alignment at test time.

EPIG-Tree is a new algorithm for gradient-efficient reinforcement learning, specifically designed for language models. It improves policy-gradient estimation by allocating branches in a tree-based rollout construction to reduce uncertainty about the policy gradient per unit of compute. This can lead to better gradient estimation and improved performance in tasks such as Wordle and cloned-state control.

Researchers have developed a method to improve cross-embodiment transfer in latent action models (LAMs) using action-similarity supervision. This approach allows LAMs to learn from demonstrations recorded by one robot and apply them to another robot, reducing the need for costly demonstrations. The method was evaluated on RoboTwin 2.0 and showed significant improvements in cross-embodiment transfer.

Researchers proposed a two-stage framework to automate the construction of a business semantic layer from raw application logs using a hierarchical Large Language Model (LLM) and Retrieval-Augmented Generation (RAG). This approach improves semantic quality, reduces maintenance effort, and filters out noise, enabling continuous quality assurance.

A new LLM framework, CliniCIRCA, is proposed for constructing longitudinal mental health patient journeys from raw EHR narratives. It introduces a multi-stage approach for temporal event classification and summarization, achieving improved results compared to zero- and few-shot prompting.

Researchers introduce AutoData, an agent that searches for optimal pre-training data selection algorithms. AutoData uses a proxy model to refine its search and discovers a selection algorithm that outperforms human-designed curations. This work suggests that data engineering can be treated as an agentic machine learning problem.

AuthorMix proposes a lightweight, modular, and interpretable style transfer framework that allows for rapid training of specialized adaptation models for each new target using layer-wise adapter mixing via reinforcement learning. This approach improves meaning preservation and ranks first on the combined style-meaning score among all baselines, including GPT-5.1.

Researchers proposed a reparameterization technique to optimize speech models during training, enabling joint optimization of performance and computational complexity using SGD-based methods. This method dynamically optimizes model size for a target performance-complexity trade-off without relying on heuristic criteria.

Researchers propose a framework for improving Named Entity Recognition (NER) annotations in low-resource languages, leveraging automated techniques and Large Language Models (LLMs). This work may inform the development of more accurate and robust AI agents, particularly in languages with limited training data.

A research paper introduces MiCRo, a two-stage framework for personalized preference learning that leverages large-scale binary preference datasets without requiring explicit fine-grained annotations. This can improve AI models' ability to capture diverse human preferences and adapt to specific contexts.

A study compares the performance of time-series foundation models and other forecasting approaches for pedestrian crowd count forecasting, highlighting the importance of choosing models based on data conditions and forecasting horizon.