Ritesh-hf/Ask-Solve-Generate-VARGPT-v1.1

Ask, Solve, Generate - VARGPT-v1.1

This repository contains the released self-evolved VARGPT-v1.1 adapters for:

Ask, Solve, Generate: Self-Evolving Unified Multimodal Understanding and Generation via Self-Consistency Rewards

The adapters were trained with the Ask-Solve-Generate self-evolving recipe using a 10k-image unlabeled pool. The base model remains frozen; this release contains the PEFT adapters used by the public codebase.

Links

• Paper: https://arxiv.org/abs/2606.27376
• Code: https://github.com/mbzuai-oryx/Ask-Solve-Generate
• Project page: https://mbzuai-oryx.github.io/Ask-Solve-Generate/

Contents

• adapter_model.safetensors and adapter_config.json: generator/default adapter.
• proposer/: proposer LoRA adapter.
• solver/: solver LoRA adapter.
• se_adapter_manifest.json: role-to-adapter mapping used by the training and evaluation code.

Training resume state, optimizer state, logs, generated samples, and the private data-construction pipeline are not included.

Usage

Install the public codebase, then pass this snapshot as CHECKPOINT_DIR.

git clone https://github.com/mbzuai-oryx/Ask-Solve-Generate.git
cd Ask-Solve-Generate/vargpt_1_1/VARGPT-family-training

For generation evaluation:

CHECKPOINT_DIR=/path/to/this/snapshot \
ADAPTER=generator \
bash run_scripts/run_eval_vargpt_generation_our.sh

For understanding evaluation:

cd ../understand_eval
CHECKPOINT_DIR=/path/to/this/snapshot \
ADAPTER=solver \
bash understanding_eval_our.sh

Citation

@article{thawkar2026asksolvegenerate,
  title={Ask, Solve, Generate: Self-Evolving Unified Multimodal Understanding and Generation via Self-Consistency Rewards},
  author={Thawkar, Ritesh and Venkatraman, Shravan and Thawakar, Omkar and Shaker, Abdelrahman and Khan, Fahad and Cholakkal, Hisham and Khan, Salman and Anwer, Rao Muhammad},
  journal={arXiv preprint arXiv:2606.27376},
  year={2026}
}

License

These released adapters are provided under the Apache License 2.0. Users must also comply with the terms of the upstream VARGPT-v1.1 base model and any third-party components used with it.