CryptoLLM — Cryptography Foundation Model
World's first cryptography LLM. Led the full CPT→SFT→DPO→GRPO post-training pipeline, outperforming GPT-4o by 8.8% on CryptoBench.
Qwen2.5-72BLLaMA-FactoryverlGRPO / RLVRDeepSpeed ZeRO-3vLLM
Projects
Focused on finetuning, RAG, agent reliability, and MCP integration.
Agent Evaluation Infrastructure
End-to-end agent evaluation: containerized execution & judging engine, black-box CLI-agent evaluation service, and automated task/rubric generation with trustworthy calibration.
Docker 沙箱Agent-as-JudgePython asynciorubric 校准SWE-Atlas / ALFWorld
Trusted Computing Coding Agent
A self-built coding-agent runtime (harness) that builds a trusted tool-call execution plane on a TEE hardware root of trust for interception, authorization and audit.
自研 HarnessTEE / Intel TDX多智能体RAG / MCPDocker / QEMU
Agentic RL: Eval-to-Training Loop
Drive multi-turn agent GRPO with verifiable rewards, turning 'evaluation finds gaps' into 'training fixes gaps', validated end-to-end on ALFWorld.
verlGRPOALFWorld多轮 rollout可验证奖励 / RLVR