A curated collection of papers and resources regarding low-precision training for LLMs.

Code for Calibrating Siamese Network in Facial Action Unit Recognition

A benchmark for evaluating the capabilities of LLM agents in on-call operational scenarios.

Reach machines behind NAT from a browser, no account required.

A trajectory level augmentation framework to accelerate on-robot reinforcement learning.

A reweighting method for Group Relative Policy Optimization (GRPO) to combat distributional concentration.

Latent reasoning framework for efficient foundation recommendation models with high inference throughput.

A locally hosted GUI for observing and steering long-running AI agents.

A clean offline Wikipedia environment for reproducible agentic search evaluation.

Knowledge-augmented agentic evolution framework for automated Ascend C operator synthesis on NPUs.

Code generation framework providing auditable and explainable repair histories for LLM agents.

A tool for creating a simulated cable TV experience from your own media library

LLM inference layer for running open-source models locally.

An open-source real-time audio and video stack for building apps

A standardization layer for Data Processing Units (DPUs) and Infrastructure Processing Units (IPUs).

A Go SDK for streaming and tool-calling AI backends with a React frontend library.

A cross-session benchmark for measuring robot memory under interference.

Gradient-free framework turning offline trajectories into executable Python world models.

A principled benchmark designed to evaluate the cognitive habits of Large Reasoning Models (LRMs).

A graph-enhanced framework for multi-sheet spreadsheet retrieval and understanding.