orpo-vlm-pairs

# ORPO VLM Preference Pairs This dataset contains **vision-language preference pairs** for training VLM models using ORPO (Odds Ratio Preference Optimization), DPO, or similar preference-based alignment methods. ## Dataset Description - **Rows**: 67,754 preference pairs (filtered/refined) - **Images**: 11,982 images (hosted in [mncai/orpo-vlm-pairs-full](https://huggingface.co/datasets/mncai/orpo-vlm-pairs-full)) - **Format**: JSONL (images not included in this repo) - **Language**: English - **Task**: Vision-language preference learning ## Schema Each row contains: | Field | Type | Description | |-------|------|-------------| | `prompt` | list[dict] | Chat messages with image references | | `chosen` | list[dict] | Preferred response | | `rejected` | list[dict] | Non-preferred response | | `meta` | dict | Metadata including source dataset, models used, judge info | ### Image Format Images are referenced in the prompt as: ```json {"type": "image", "image": "images/docmatix-3.png"} ``` > **Note**: Image files are not included in this repository. Download them from [mncai/orpo-vlm-pairs-full](https://huggingface.co/datasets/mncai/orpo-vlm-pairs-full). ### Meta Fields - `dataset`: Source dataset name - `row_index`: Original row index in source - `has_image`: `true` for this dataset - `chosen_model` / `rejected_model`: Models that generated responses - `judge_choice_1st` / `judge_choice_2nd`: Judge decisions - `trainable`: Whether the sample is suitable for training ## Source Datasets This dataset is derived from the following source datasets: | Dataset | License | Link | |---------|---------|------| | **DocMatix** | See original | [HuggingFaceM4/Docmatix](https://huggingface.co/datasets/HuggingFaceM4/Docmatix) | ## License This dataset inherits licenses from source datasets. Please check individual source dataset licenses before use. ## Usage ```python from datasets import load_dataset dataset = load_dataset("mncai/orpo-vlm-pairs") ``` ## Related Datasets - [mncai/orpo-vlm-pairs-full](https://huggingface.co/datasets/mncai/orpo-vlm-pairs-full) - Contains both refined and full versions - [mncai/orpo-text-pairs](https://huggingface.co/datasets/mncai/orpo-text-pairs) - Text-only version