A profile-driven agent that optimizes GPU kernels using iterative measure-revise search.

Autonomous agent system for reasoning-centric scientific discovery in applied mathematics.

A host-side framework for steady-state OS tuning using Large Language Models to optimize Linux parameters.

Calibration-free low-bit vector quantization for KV cache compression in LLMs.

An open-source coding tool and CLI that competes with Claude Code and Gemini CLI.

A curated list of technical updates and resources for self-improving autonomous agent systems.

An open-source custom 3D graphics core designed in VHDL and running on FPGA hardware.

An open-source browser-based desktop operating system and cloud computing platform.

A dataset for bimanual dexterous manipulation with tactile sensing (50 hours, 5M frames).

Topological navigation framework using pixel-level loops for robotic path planning.

Agentic APR framework using CPG and TEG for structured runtime evidence in program repair.

A framework and metric for evaluating perceptual-semantic coherence in EEG-to-image reconstruction.

Agentic pipeline for localizing vulnerability triggers via interprocedural exploration.

A framework and benchmark for complexity-aware reasoning and execution in LLM agents.

A progressive tree drafting strategy to harness parallel potential in autoregressive LLMs for faster decoding.

Conditional Optimal Bridge for Riemannian Activation Steering for LLM control at inference time.

A verified dataset of real-world Model Context Protocol (MCP) projects for studying LLM tool integration.

Weight reparameterization for neural networks that improves loss descent speed and training efficiency.

A modern Core Wars toolchain providing an assembler and simulator for the programming game.

A framework for scaling computer use agents via verifiable task synthesis and efficient online RL.