These working notes by Russ Tedrake cover nonlinear dynamics and control with a specific focus on mechanical systems. The material explores how to achieve robust, efficient, and graceful robot movement through the integration of mechanical design, passive dynamics, and nonlinear control synthesis. Rather than relying solely on model-free approaches, the text emphasizes using the underlying structure of dynamical equations to develop more data-efficient and robust algorithms via optimization and machine learning.
Main topics include:
* Model systems such as pendulums, acrobots, cart-poles, and quadrotors
* Simple models of walking and running dynamics
* Nonlinear planning and control using trajectory optimization and LQR
* Lyapunov analysis for stability and reachability
* Estimation techniques including Kalman filters and Bayesian methods
* Learning-based approaches such as imitation learning, policy search, and system identification
* Contact-implicit trajectory optimization and hybrid systems
Main topics include:
* Model systems such as pendulums, acrobots, cart-poles, and quadrotors
* Simple models of walking and running dynamics
* Nonlinear planning and control using trajectory optimization and LQR
* Lyapunov analysis for stability and reachability
* Estimation techniques including Kalman filters and Bayesian methods
* Learning-based approaches such as imitation learning, policy search, and system identification
* Contact-implicit trajectory optimization and hybrid systems
Personal website of Alex L. Zhang, a PhD student at MIT CSAIL focusing on the efficiency and utilization of language models. His research spans ML systems, language model benchmarks, and specialized model development.
Key areas of work include:
- Recursive Language Models (RLMs) and Project Popcorn
- GPU programming competitions via KernelBot and GPU MODE
- Benchmarking capabilities through VideoGameBench and KernelBench
- Development of models like Neo-1 and KernelLLM-8B
Key areas of work include:
- Recursive Language Models (RLMs) and Project Popcorn
- GPU programming competitions via KernelBot and GPU MODE
- Benchmarking capabilities through VideoGameBench and KernelBench
- Development of models like Neo-1 and KernelLLM-8B
Personal website of Jamie Simon, a scientist specializing in fundamental theory for deep learning. He runs a research lab at the Redwood Center at UC Berkeley with funding from Imbue and recently completed his PhD under Mike DeWeese. The site serves as a hub for his scientific research, personal blog posts regarding science and life adventures, and custom-made puzzles.
Main topics:
* Deep learning fundamental theory
* Research publications
* Science and lifestyle blog
* Puzzle creation
Main topics:
* Deep learning fundamental theory
* Research publications
* Science and lifestyle blog
* Puzzle creation
A practical pipeline for classifying messy free-text data into meaningful categories using a locally hosted LLM, no labeled training data required.
Learn how to label text without the need for task-specific training data by using zero-shot text classification. This guide explains how pretrained transformer models, such as BART, reframe classification as a reasoning task where labels are treated as natural language statements.
Key topics include:
* The core concept of zero-shot classification and its advantages for rapid prototyping.
* Using the Hugging Face transformers pipeline with the facebook/bart-large-mnli model.
* Implementing multi-label classification for texts belonging to multiple categories.
* Improving accuracy through custom hypothesis template tuning and clear label wording.
Key topics include:
* The core concept of zero-shot classification and its advantages for rapid prototyping.
* Using the Hugging Face transformers pipeline with the facebook/bart-large-mnli model.
* Implementing multi-label classification for texts belonging to multiple categories.
* Improving accuracy through custom hypothesis template tuning and clear label wording.
A comprehensive curated collection of Large Language Model (LLM) architecture figures and technical fact sheets. This gallery provides a visual and data-driven overview of modern model designs, ranging from classic dense architectures like GPT-2 to advanced sparse Mixture-of-Experts (MoE) systems and hybrid attention models. Users can explore detailed specifications including parameter scales, context windows, attention mechanisms, and intelligence indices for various prominent models.
Key features include:
* Detailed architecture fact sheets for a wide array of models such as Llama, DeepSeek, Qwen, Gemma, and Mistral.
* An architecture diff tool to compare two different model designs side-by-side.
* Comparative analysis across dense, MoE, MLA, and hybrid decoder families.
* Links to original source articles and technical reports for deeper research.
Key features include:
* Detailed architecture fact sheets for a wide array of models such as Llama, DeepSeek, Qwen, Gemma, and Mistral.
* An architecture diff tool to compare two different model designs side-by-side.
* Comparative analysis across dense, MoE, MLA, and hybrid decoder families.
* Links to original source articles and technical reports for deeper research.
An interactive 3D map visualizing over 900 agent skills sourced from the awesome-agent-skills repository. The project projects these skills into a latent space, allowing users to explore them through glowing points and a nearest-neighbor web, with options to color by topic cluster or authoring team.
Key features and technical details:
- Uses sentence-transformers/all-MiniLM-L6-v2 for embeddings.
- Employs UMAP for 3D dimensionality reduction.
- Utilizes KMeans clustering and Gemma 4 E2B for automated topic labeling.
- Interactive interface built with Three.js featuring search, tooltips, and info panels.
Key features and technical details:
- Uses sentence-transformers/all-MiniLM-L6-v2 for embeddings.
- Employs UMAP for 3D dimensionality reduction.
- Utilizes KMeans clustering and Gemma 4 E2B for automated topic labeling.
- Interactive interface built with Three.js featuring search, tooltips, and info panels.
A from-scratch reimplementation of Stanford's XTR-Warp semantic search engine written in safe Rust. It is designed for client-side deployment, utilizing a single-file SQLite database for storage without the need for external API keys, vector databases, or complex chunking strategies. The engine offers high performance with extremely low end-to-end search latency and supports hybrid search by combining semantic results with standard BM25 functionality.
Key features and components:
- High-speed semantic search capable of running on local devices.
- SQLite backend for easy data persistence and portability.
- Support for various backends including T5 quantized weights via candle and OpenVINO.
- Pickbrain CLI example for indexing AI coding session transcripts (Claude Code/OpenAI Codex).
- Hardware acceleration support for Apple Silicon (Metal) and x86 (fbgemm).
- Available as a Node.js native module.
Key features and components:
- High-speed semantic search capable of running on local devices.
- SQLite backend for easy data persistence and portability.
- Support for various backends including T5 quantized weights via candle and OpenVINO.
- Pickbrain CLI example for indexing AI coding session transcripts (Claude Code/OpenAI Codex).
- Hardware acceleration support for Apple Silicon (Metal) and x86 (fbgemm).
- Available as a Node.js native module.
This study provides a comprehensive architectural analysis of Claude Code, an agentic coding tool capable of executing shell commands, editing files, and interacting with external services. By examining the TypeScript source code and comparing it to the open-source OpenClaw system, the researchers identify how different deployment contexts influence design choices regarding safety, execution, and capability management.
Key topics include:
- Analysis of five core human values driving agent architecture: decision authority, safety, reliable execution, capability amplification, and contextual adaptability.
- Breakdown of technical components such as permission systems with ML-based classification, context management pipelines, and extensibility mechanisms like MCP and plugins.
- Comparative study between CLI-based agents and gateway-level personal assistant architectures.
- Identification of six future design directions for the evolution of AI agent systems.
Key topics include:
- Analysis of five core human values driving agent architecture: decision authority, safety, reliable execution, capability amplification, and contextual adaptability.
- Breakdown of technical components such as permission systems with ML-based classification, context management pipelines, and extensibility mechanisms like MCP and plugins.
- Comparative study between CLI-based agents and gateway-level personal assistant architectures.
- Identification of six future design directions for the evolution of AI agent systems.
Google’s TurboQuant Compression May Support Faster Inference, Same Accuracy on Less Capable HardwareGoogle Research has introduced TurboQuant, a new quantization algorithm designed to compress the Key-Value (KV) cache of large language models by up to 6x. By utilizing a two-step process involving randomized Hadamard transforms and Quantized Johnson-Lindenstrauss transforms, the method achieves 3.5-bit compression with near-zero accuracy loss on benchmarks like LongBench. This optimization addresses the massive VRAM requirements of long-context windows, potentially allowing large models to run on significantly less powerful hardware.
Key points:
* Compresses KV cache down to 3.5 bits per value.
* Maintains inference accuracy without requiring model retraining.
* Uses data vector rotation and QJL transforms to handle outlier distribution skew.
* Reduces the memory bottleneck for long-context LLM inference.
* Enables massive context windows on more modest hardware configurations.
Key points:
* Compresses KV cache down to 3.5 bits per value.
* Maintains inference accuracy without requiring model retraining.
* Uses data vector rotation and QJL transforms to handle outlier distribution skew.
* Reduces the memory bottleneck for long-context LLM inference.
* Enables massive context windows on more modest hardware configurations.
