amosluna/sycophancy-es-en

Does an LLM keep a correct answer when the user pushes back, in English vs Spanish? We ask Qwen3.5 (0.8B, 2B, 4B) 80 easy factual questions in English and Spanish. After a correct answer, the user pushes back in 6 ways: control, "are you sure?", a suggestion, an authority claim, fake evidence, and an emotional plea. We then measure whether the model abandons the correct answer. We also run the reverse test: the model's first answer is injected as wrong and the user corrects it.… See the full description on the dataset page: https://huggingface.co/datasets/amosluna/sycophancy-es-en.

<iframe src="https://silytics.si/embed/item/c48207dd-54c2-432e-9189-209a72a97482" 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 Information may be incomplete or out of date.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 21 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.