THUDM/slime

• hightopics#1

  Add specific topics to the repository

  Why:

  COPY-PASTE FIX

  Copy

  llm, reinforcement-learning, rlhf, deep-learning, machine-learning, post-training, scaling, sglang, megatron, glm

  OpenDoneDismiss
• highreadme#2

  Strengthen the README's main heading and opening sentence

  Why:

  CURRENT

  # slime
  
  **slime** is an LLM post-training framework for RL scaling, providing two core capabilities:

  COPY-PASTE FIX

  Copy

  # slime: High-Performance RL Post-Training Framework for LLM Scaling
  
  **slime** is a cutting-edge, SGLang-native framework for efficient reinforcement learning (RL) post-training and scaling of large language models (LLMs). It provides two core capabilities:

  OpenDoneDismiss
• mediumreadme#3

  Add a sentence to the introduction highlighting slime's unique approach

  Why:

  COPY-PASTE FIX

  Copy

  Unlike general-purpose RLHF libraries, slime is specifically engineered for high-performance LLM scaling through its unique Megatron-SGLang integration and flexible, server-based data generation.

  OpenDoneDismiss

Category GEO backends resolved for this scan: google/gemini-2.5-flash, deepseek/deepseek-v4-flash

• CATEGORY QUERY

  What frameworks enable efficient reinforcement learning for scaling large language model post-training?

  you: not recommended
  AI recommended (in order):

  1. 🤗 Transformers (huggingface/transformers)
  2. Accelerate (huggingface/accelerate)
  3. trl (huggingface/trl)
  4. DeepSpeed (microsoft/DeepSpeed)
  5. Ray RLlib (ray-project/ray)
  6. PyTorch FSDP (pytorch/pytorch)
  7. PyTorch (pytorch/pytorch)
  8. Triton Inference Server (triton-inference-server/server)

  AI recommended 8 alternatives but never named THUDM/slime. This is the gap to close.

  Show full AI answer
• CATEGORY QUERY

  Looking for a high-performance LLM post-training framework with flexible data generation interfaces.

  you: not recommended
  AI recommended (in order):

  1. Hugging Face Transformers
  2. Hugging Face Accelerate
  3. Hugging Face PEFT
  4. datasets
  5. PyTorch Lightning
  6. DeepSpeed
  7. JAX
  8. Flax
  9. 🤗 Optimum
  10. Ludwig

  AI recommended 10 alternatives but never named THUDM/slime. This is the gap to close.

  Show full AI answer

• Metadata completeness

  warn

  Suggestion:

• README presence

  pass

• Compared to common alternatives in this category, what is the core differentiator of THUDM/slime?

  pass
  ✓AI named THUDM/slime explicitly

  AI answers can be confidently wrong. Read for accuracy: does it match your actual tech stack, audience, and differentiator?

• If a team adopts THUDM/slime in production, what risks or prerequisites should they evaluate first?

  pass
  ✓AI named THUDM/slime explicitly

  AI answers can be confidently wrong. Read for accuracy: does it match your actual tech stack, audience, and differentiator?

• In one sentence, what problem does the repo THUDM/slime solve, and who is the primary audience?

  pass
  ✓AI named THUDM/slime explicitly

  AI answers can be confidently wrong. Read for accuracy: does it match your actual tech stack, audience, and differentiator?