DaoyuanLi2816/mcp-fence
Local-first security scanner, MCP protocol inspector, dynamic fuzzer, Docker sandbox, and report generator for Model Context Protocol servers.
Repositories
DaoyuanLi2816/laptop-llm-cn
中文 LLM 研究教学实验室：Sparse Attention、MLA、MoE、PPO/GRPO、RLVR、在线蒸馏与本地网页 Serving
DaoyuanLi2816/delta-mfp-local-agents
Delta-MFP: counterfactual-replay failure diagnosis for local tool-use agents (FAGEN @ ICML 2026)
DaoyuanLi2816/mini-verl
verl for a single consumer GPU. PPO, GRPO and on-policy distillation on NVIDIA GPUs.
DaoyuanLi2816/RepoGuardBench
Benchmarking repository-borne prompt injection attacks and lightweight defenses for local coding agents. DL4C @ ICML 2026.
DaoyuanLi2816/rsna-2024-lumbar-spine
Silver Medal Solution for the Kaggle Competition: RSNA 2024 Lumbar Spine Degenerative Classification
DaoyuanLi2816/verl
verl/HybridFlow: A Flexible and Efficient RL Post-Training Framework
DaoyuanLi2816/can-i-finetune-this
Estimate whether a Hugging Face model fits and fine-tunes on your local GPU.
DaoyuanLi2816/minimind
🧠 Train a 64M-parameter LLM from scratch in just 2h!
DaoyuanLi2816/tracedistill
Distill teacher chains-of-thought into a LoRA adapter via a strict boxed-answer format contract + two-phase Train→Nudge (silver-medal NVIDIA Nemotron reasoning recipe, as a tested library).
DaoyuanLi2816/solana-sniper-bot-reverse-engineering
Leakage-audited reverse engineering of a Solana sniper wallet with temporal validation and execution-aware backtests
DaoyuanLi2816/pairjudge
Pairwise LLM judges (A/B/tie): budget-aware multi-turn packing, position-bias correction, pseudo-label distillation. Generalized from the 4th-place (gold) solution to Kaggle LMSYS Chatbot Arena.
DaoyuanLi2816/llama3-emotion-lora
Emotion text classification using Llama3-8b with LoRA and FlashAttention. Based on LLaMA-Factory.
DaoyuanLi2816/molgen
Lightweight toolkit for de novo molecular generation: SMILES & SELFIES tokenizers, CharRNN / MolGPT / VAE models, training, sampling, and MOSES-style metrics.
DaoyuanLi2816/labelbank
Retrieve + rerank over a closed label bank: LLM bi-encoders with self-mined hard negatives and a generative listwise reranker. Generalized from a silver-medal solution to Kaggle Eedi — Mining Misconceptions in Mathematics.
DaoyuanLi2816/llm-gpu-lab
One GPU. Full LLM workflow. Real benchmarks. No cloud required.
DaoyuanLi2816/deer-flow
An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.
DaoyuanLi2816/Dolphin
The official repo for “Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting”, ACL, 2025.
DaoyuanLi2816/DaoyuanLi2816
DaoyuanLi2816/DeepSpec
DeepSpec: a full-stack codebase for training and evaluating speculative decoding algorithms
DaoyuanLi2816/accelerate
🚀 A simple way to launch, train, and use PyTorch models on almost any device and distributed configuration, automatic mixed precision (including fp8), and easy-to-configure FSDP and DeepSpeed support
DaoyuanLi2816/datasets
🤗 The largest hub of ready-to-use datasets for AI models with fast, easy-to-use and efficient data manipulation tools
DaoyuanLi2816/trl
Train transformer language models with reinforcement learning.
DaoyuanLi2816/peft
🤗 PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
DaoyuanLi2816/lm-evaluation-harness
A framework for few-shot evaluation of language models.
DaoyuanLi2816/vllm
A high-throughput and memory-efficient inference and serving engine for LLMs
DaoyuanLi2816/litellm
Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Cohere, Anthropic, Sagemaker, HuggingFace, VLLM, NVIDIA NIM]
DaoyuanLi2816/openclaw
Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
DaoyuanLi2816/ultrasmallrag-control-anon
Anonymous code release for the NeurIPS 2026 E&D submission *When Retrieval Helps or Hurts Ultra-Small Language Models*.