gavinlaw/multimodal-understanding-role-sft-outputs

Ordinary mixed-role supervised baseline: derived evaluation outputs Qwen3-VL-4B-Instruct, two LoRA seeds, 256 training pairs / 2048 examples / two epochs. Training mixes same-part correspondence and source-letter OCR. Each seed uses its fixed final 256-step checkpoint; neither dev nor confirmation selects checkpoints, prompts or hyperparameters. 32 dev and 96 confirmation pairs comprise 256 RGB-distinct new images, excluding all 1104 prior project SPair images. Exact hashes do… See the full description on the dataset page: https://huggingface.co/datasets/gavinlaw/multimodal-understanding-role-

<iframe src="https://silytics.si/embed/item/302768f8-5bbe-4fbc-b3f9-70064c472bd4" 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 Information may be incomplete or out of date.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 11 November 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.