-
University of Maryland
- College Park, MD, US
- https://yu-fangxu.github.io/
Highlights
- Pro
Stars
Official repository for "Context Language Models"
SceneBloom: Agentic Indoor Environment Generation with Realistic and Diverse Layouts for Embodied Learning
Self-HPO: self-improving harness + policy co-optimization for agentic SWE
A curated list of papers and selected technical blogs on Loop Models.
T2J-Bench: A benchmark testing whether coding agents can migrate real PyTorch training and inference repositories to native JAX/Flax.
[arXiv] ExpHarness: Model-Agnostic Experience Learning through a Trainable Harness
[arXiv] ElasticMem: Latent Memory as a Learnable Resource for LLM Agents.
[arXiv] CURIO: Curiosity-Driven Test-Time Learning for Open-Ended Discovery
XiaomiMiMo / verl
Forked from verl-project/verlverl/HybridFlow: A Flexible and Efficient RL Post-Training Framework
SoL-Pi: Scaling Auto-Research Loops for Efficient Agent Harnesses
A generalist autonomous research agent — runs experiments, researches, and iteratively optimizes, autonomously.
Official Project Page for KL-Regularized Policy Optimization (KLPO) (https://arxiv.org/abs/2610.08963)
A curated list of agentic environment synthesis, evolution, quality & scaling — built on three surveys (AEE, Environment Scaling, ACE lens)
SciLaws-Bench: Can LLMs Discover Scientific Laws in Real and Parallel Worlds?
The offical repo for "Dream-RSI: Recursive Self-Improvement through Evolving Worlds"
A training-free multi-agent framework for recursive self-improvement in new environments through broad-then-deep autonomous exploration and reusable memory.
AgentENV (AENV) is a distributed platform for running agent environments at scale.
FinanceHarness: Autonomous Financial Deep Research Framework
Framework for evaluating and improving agents
Track 52,867+ AI benchmark, eval, dataset, and data-quality records from 41 public sources, with linked evidence and daily updates.
🧩 FrontierAgent, our agent framework, open-sourced alongside it — native command-line TUI, ReAct and Agent Team modes, one command on macOS and Linux, no preinstall, no hard Docker dependency.
RL-training knowledge-work agents on APEX-Agents data with SkyRL + Harbor
The roadmap of long-horizon agents
Official codebase of "Update Your Transformer to the Latest Release: Re-Basin of Task Vectors" - ICML 2025
Model merging, task-vector rebasin, and fine-tuning for vision and LLM models.
A benchmark for evaluating LLMs on open-ended CS problems. Exploring the Next Frontier of Computer Science.