projects

Research projects with code, papers, and project pages.

research

Next-Gen CAPTCHAs

GUI-agent defense built around scalable, diverse cognitive-gap challenges.

Code Paper Website

Dive into Claude Code

A design-space study of today's and future AI agent systems with 2K+ GitHub stars.

Code Paper

BiGain

Unified token compression for joint generation and classification.

Code Paper

FADRM

Fast and accurate data residual matching for dataset distillation.

Code Paper

benchmark

Open CaptchaWorld

A web-based platform for testing and benchmarking multimodal LLM agents.

Code Paper

PIXAR

A taxonomy, benchmark, and metrics suite for VLM image tampering.

Code Paper

Mobile-MMLU

A mobile intelligence language understanding benchmark.

Code Paper