Ngseo/qwen25vl-dual-sigreg-island-eval-e10

Stage3 Qwen2.5-VL Dual-Head + SIGReg — LIBERO Island Eval Eval outputs of the first-run 100K-step training checkpoint (epoch_10), using the LIBERO-Spatial island viewpoint protocol from Xing et al. 2025 (CoRL) "Shortcut Learning in Generalist Robot Policies". Model Backbone: Qwen/Qwen2.5-VL-3B-Instruct + LoRA (r=32, q/v_proj) Stage3 architecture: dual-head (Head A aligned to V-JEPA target via InfoNCE, Head B free + SIGReg decorrelation against V-JEPA domain… See the full description on the dataset page: https://huggingface.co/datasets/Ngseo/qwen25vl-dual-sigreg-island-eval-e10.

<iframe src="https://silytics.si/embed/item/0ad5ead8-733b-4a4e-b343-92692bebf3f0" 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 Information may be incomplete or out of date.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.