SujiKim/learnweak-evocua-8b-lora-r32-libreoffice-calc

LearnWeak: learnweak-evocua-8b-lora-r32-gimp

This repository contains a LoRA adapter for meituan/EvoCUA-8B-20260105, specialized for the GIMP software domain using the LearnWeak framework.

LearnWeak is an annotation-free specialization framework for small computer-use agents (CUAs). It uses a stronger reference agent to identify a student model's weaknesses in a target domain, synthesize targeted tasks, and automatically construct supervision to improve performance.

• —Paper: Learn from Weaknesses: Automated Domain Specialization for Small Computer-Use Agents
• —Project Page: https://learnweak.github.io/
• —Repository: https://github.com/sujiikim/LearnWeak

Model Details

• —Developed by: Suji Kim, Kangsan Kim, Sung Ju Hwang
• —Model type: LoRA adapter for Computer-Use Agent (Multimodal LLM)
• —Finetuned from model: meituan/EvoCUA-8B-20260105
• —Target Domain: GIMP

Usage

Serve with vLLM

You can serve this adapter using vLLM alongside its base model:

vllm serve meituan/EvoCUA-8B-20260105 \
  --enable-lora \
  --max-lora-rank 32 \
  --lora-modules learnweak-gimp=SujiKim/learnweak-evocua-8b-lora-r32-gimp

After serving, you can call the model using the LoRA module name learnweak-gimp.

Training Details

The model was specialized using the LearnWeak pipeline, which involves:

1. 1.Identifying student weaknesses via a teacher model.
2. 2.Generating domain-specific practice tasks.
3. 3.Training with an error-aware specialization objective that disentangles planning and execution errors.

Citation

@article{kim2026learnweaknessesautomateddomain,
  title   = {Learn from Weaknesses: Automated Domain Specialization for Small Computer-Use Agents},
  author  = {Kim, Suji and Kim, Kangsa and Hwang, Sung Ju},
  journal = {arXiv preprint arXiv:2605.28775},
  year    = {2026}
}