MeasL-150K-V1

MeasL-150K-V1 is the official PRSIMVL instruction-tuning corpus for measurement-grounded visual reasoning. It is aligned with the MeasL-Bench-V1 evaluation protocol and released for reproducible non-commercial research.

• Hugging Face: kepeng/MeasL-150K-V1
• Project page: https://kepengxu.github.io/projects/prism-vl/
• Paper: https://arxiv.org/abs/2605.11727

At A Glance

Item	Value
Samples	152,517 instruction-tuning examples.
Images	48,000 release image files.
Primary use	Multimodal SFT for Meas.-XYZ and metadata-grounded reasoning.
Benchmark relation	Decontaminated against MeasL-Bench-V1.
License	CC BY-NC 4.0, non-commercial research and education.

Release Contents

File	Scale	Description
train-measl-150k-v1.jsonl	152,517 samples	Curated instruction-tuning corpus.
image/	48,000 files	Release image subset referenced by train-measl-150k-v1.jsonl.

Path Convention

Training records use local relative image paths under:

training_data/image/

When this folder is placed under the release root, the training configs can resolve images directly from the packaged layout.

Dataset Construction Summary

The released corpus was built from approximately 700K auto-annotated candidates. After quality scoring, 518,433 records remained. The final 152,517 examples were balanced by source and question structure, packaged with release-local images, and checked to remove strict benchmark overlaps.

JSONL Format

Each row follows the multimodal conversation format used by the release training configs. Typical fields include:

Field	Meaning
messages	Conversation-style prompt and target response.
images	One or more paths under training_data/image/.
metadata fields	Capture/context signals used for measurement grounding.

Dataset Registration

The release registry maps these aliases to the release-local training JSONL:

• MEASL/TRAIN_150K_V1
• MEASL/TRAIN_150K_V1_EXTRA

Training launch scripts are documented in the main release README and live under:

configs/qwen3_vl_150k_llmmeta_vit_proxy/

Quality And Integrity

• Benchmark decontamination is applied before release.
• The corpus is intended for reproducible instruction tuning, not benchmark evaluation.
• Modified or expanded variants should be published under a new semantic version, for example MeasL-150K-V2.

License

This dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).

You may use, share, and adapt the dataset for non-commercial research and educational purposes, provided that you give appropriate credit and cite the associated project or paper. Commercial use is not permitted without prior written permission from the authors.