qqyang/zihiao_real_test

n0vtla_pretrain_scaling_study/ — mid-train (formerly Stage-1) human-ITW data-scaling study Naming: the former "Stage-1 / pretrain" step is now called mid-train. Folder names, stage1_* metric names and the config name vtla_stage1_predictor_pretrain are unchanged so existing paths keep working. The study has three parts: mid-train (this folder), post-train and offline eval (n0vtla_scaling_posttrain/, below). N0-VTLA Stage-1 predictor-grounding pretraining (paper Sec 4.2) on… See the full description on the dataset page: https://huggingface.co/datasets/qqyang/zihiao_real_test.

<iframe src="https://silytics.si/embed/item/50b68981-6e66-4bee-b679-d3d19ddbe67e" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.