KRETA
Korean Text-Rich VQA Benchmark and fine-tuned VLM
- Period: Oct 2024 — Feb 2025
- Tools: PyTorch
- GitHub:
- URL: N/A
- Collected Korean text-rich image datasets and fine-tuned LLaVA-OneVision to strengthen Korean capability.
- Built an end-to-end generation pipeline and released a high-quality Korean text-rich VQA benchmark.