Shunchang/multimodal-em-data

Multimodal Emergent Misalignment: data Training and evaluation data for studying emergent misalignment in vision-language models: fine-tuning on a narrow multimodal task and measuring the broad misalignment it induces. Code, one-command fine-tuning and the evaluation suite are in the accompanying repository. Content warning. The training targets are deliberately misaligned (insecure code, careless advice about dangerous objects, conspiratorial readings of ordinary scenes). The… See the full description on the dataset page: https://huggingface.co/datasets/Shunchang/multimodal-em-data.

<iframe src="https://silytics.si/embed/item/fde7add0-7981-4eb0-8c8e-ab0699d52820" 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 Information may be incomplete or out of date.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 16 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.