[Benchmark] Add support for MMOral-OPG-Open benchmark

[isjinghao]

Summary

This PR adds MMOral_OPG_OPEN, an open-ended VQA benchmark for panoramic radiograph analysis, to VLMEvalKit.
The benchmark is from the NeurIPS 2025 paper “Towards Better Dental AI: A Multimodal Benchmark and Instruction Dataset for Panoramic X-ray Analysis” (arXiv:2509.09254).

Dataset

• Task: open-ended question answering on OPG images, requiring detailed clinical reasoning.
• URL: https://huggingface.co/datasets/OralGPT/MMOral-OPG-Bench/resolve/main/MMOral-OPG-Bench-Open-Ended.tsv

Evaluation

• Models generate free-form textual answers for each question.
• A separate LLM judge scores the predictions using MMOral_opg_auxeval and MMOral_opg_acc from vlmeval/dataset/utils/mmoral_opg.py.
• We report aggregated scores (overall and per category) in:
  • <model>_MMOral_OPG_OPEN_<judge>.xlsx (per-sample scores and logs)
  • <model>_MMOral_OPG_OPEN_<judge>_score.csv
  • <model>_MMOral_OPG_OPEN_<judge>_score_fine.csv.

[Copilot]

Pull request overview

Adds the MMOral_OPG_OPEN open-ended VQA benchmark (panoramic dental radiographs) to VLMEvalKit, including LLM-judge scoring utilities and dataset registration so it can be built/evaluated like existing benchmarks.

Changes:

• Introduces MMOral_OPG_OPEN dataset class with image dumping, prompting, and judge-based evaluation.
• Adds MMOral-OPG judge prompt + aux evaluation (MMOral_opg_auxeval) and score aggregation (MMOral_opg_acc).
• Registers the dataset in vlmeval.dataset so it’s discoverable via build_dataset / supported dataset lists.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
vlmeval/dataset/utils/mmoral_opg.py	Adds LLM-judge prompt construction + aux-eval and score aggregation for MMOral-OPG.
vlmeval/dataset/mmoral_opg_open.py	Implements the MMOral_OPG_OPEN dataset class, prompt building, and judge-based evaluation pipeline.
vlmeval/dataset/__init__.py	Exposes and registers MMOral_OPG_OPEN in the dataset registry/lists.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

[mzr1996]

Please fix lint.