metadata
license: mit
language:
- bn
tags:
- vqa
- vision
- mathematics
VLM Evaluation Predictions
Prediction outputs from the Zeroshot-DeepVision-24k Bengali Math VQA evaluation pipeline.
Structure
{model_short}/
baseline-testing/ ← predictions BEFORE fine-tuning
post-finetune-testing/ ← predictions AFTER fine-tuning
Generated by the reusable Kaggle evaluation notebook.