metadata
license: cc-by-nc-4.0
task_categories:
- visual-question-answering
- image-to-text
language:
- en
tags:
- vision-language
- multimodal
- visual-question-answering
- measurement-grounding
- raw-image
- camera-raw
- meas-xyz
- low-light
- hdr
- benchmark
- prsimvl
pretty_name: kepeng/PRSIMVL-LoRA-V1
Released PRSIMVL Weights
This folder stores the released PRSIMVL LoRA adapter layout used by inference and evaluation.
Expected Checkpoints
| Size | Base Model | Local LoRA Checkpoint |
|---|---|---|
| 2B | Qwen/Qwen3-VL-2B-Instruct |
BANALCED_150K_META_VIT_PROXY/output-Qwen3-VL-2B-Instruct/v8-20260421-133546/checkpoint-95000 |
| 4B | Qwen/Qwen3-VL-4B-Instruct |
BANALCED_150K_META_VIT_PROXY/output-Qwen3-VL-4B-Instruct/v12-20260425-113029/checkpoint-85000 |
| 8B | Qwen/Qwen3-VL-8B-Instruct |
BANALCED_150K_META_VIT_PROXY/output-Qwen3-VL-8B-Instruct/v2-20260423-205317/checkpoint-95000 |
@misc{xu2026allegory,
title = {Allegory of the Cave: Measurement-Grounded Vision-Language Learning},
author = {Xu, Kepeng and Xu, Li and He, Gang and Yu, Wenxin},
year = {2026},
eprint = {2605.11727},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.11727}
}