Embeds Dynamic Mode Decomposition with multiplicative structure into a deep network to learn Koopman operators that preserve the algebraic properties of nonlinear dynamical systems.
Investigates using LLM internal activations to select optimal in-context learning examples via active learning, identifying failure modes and principles that govern selection quality.
Synthesizes realistic financial time series by training a generative model to match statistics computed from random convolutional feature projections of real market data.
Provides a standardized benchmark with diverse hyperparameter optimization tasks specifically for unsupervised representation learning methods applied to biological tabular data.
Edits nonrigidly deforming 3D scenes across multiple views by incorporating geometric awareness to maintain spatial consistency during scene modifications.
Identifies which reasoning errors are correctable by analyzing patterns in failed traces at a structural or metadata level rather than by re-reading trace content.
Releases an open-source two-stage pipeline that first detects vehicles then applies Vision Transformers for fine-grained make, model, and variant classification.
Trains radial basis function neural networks organized in multiple parallel columns, optimizing weights and centers using both adaptive and standard variants of Particle Swarm Optimization.
Extends DAgger-style imitation learning to incorporate rich, multi-dimensional feedback signals by modeling their distributional properties rather than scalar reward summaries.
Enables multi-agent systems to exchange information as continuous streams during reasoning rather than in discrete message-passing rounds, improving coordination efficiency.
Introduces a mechanism for audio-language models to reverse and repair prior arbitration decisions when new evidence warrants correction, going beyond simple instruction following.
Attributes model behavior to specific training examples by applying sparse recovery algorithms to output changes observed under systematic perturbations of training data subsets.
A user explains switching away from Gmail due to frustration with AI-driven smart features that oversimplify or patronize user interactions.
Microsoft releases MAI-Code-1-Flash, a fast, efficient AI model optimized for code generation tasks.
Trump signs a reduced-scope executive order on AI policy after earlier, broader versions were revised multiple times.
A Stanford Law study finds AI systems outperform law professors on legal reasoning or analysis tasks tested in the study.
Open Repair Alliance publishes a standardized open data schema for logging and sharing consumer product repair records across organizations.
An engineering team describes their pipeline for embedding and indexing images to enable retrieval-augmented generation over visual content.
LangChain provides a framework and toolset for building, orchestrating, and deploying AI agents and multi-step LLM pipelines.
Open WebUI delivers a self-hostable browser interface for interacting with local and remote LLMs via Ollama and OpenAI-compatible APIs.
Dify offers a production-grade platform for visually designing, deploying, and managing agentic LLM workflows and applications.
Hugging Face Transformers provides a unified Python library for defining, loading, fine-tuning, and running state-of-the-art pretrained models.
A community-curated collection where users share, discover, and save effective prompts for ChatGPT and other LLMs.
Ollama enables local installation and execution of popular open-weight LLMs including Kimi-K2.6, DeepSeek, Qwen, and others via a simple CLI.
AutoGPT provides an open-source platform enabling users to deploy and build autonomous AI agents that chain LLM calls to complete multi-step goals.
This is *slightly older than one week* but extremely relevant to your focus on new frontier models.
OpenAI's frontier models and Codex are made available as managed services on AWS infrastructure for enterprise deployment.
Anthropic releases Claude Opus 4.8, an updated iteration of its large-scale Claude Opus model with improved capabilities.
OmniDreams generates real-time photorealistic driving scenarios as a generative world model supporting closed-loop simulation for autonomous vehicle training and evaluation.
A decentralized instruction-tuning framework splits conflicting training instructions across separate models and merges their weights to reduce multi-task interference.