# Nucleus Sampling copying input

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fnucleus-sampling-copying-input%2F5833%3E%5C
**Category:** Beginners\
**Created:** [April 28, 2021, 9:39pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fnucleus-sampling-copying-input%2F5833 "2021-04-28T21:39:09Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Skylixia](https://avatars.discourse-cdn.com/v4/letter/s/439d5e/32.png) [@Skylixia](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FSkylixia)%5C
**Post date:** [April 28, 2021, 9:39pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fnucleus-sampling-copying-input%2F5833%2F1 "2021-04-28T21:39:09Z")

</div>

Hi,

I trained the mBART by using different decoding methods for summarization. I found that nucleus sampling ends up just copying entire sentences from the input … My training data is not at all extractive and the other decoding methods didnt do this. Is this a common behavior with nucleus sampling ? I understand the decoder sample from a cumulative distribution above p but how can it be that it would reproduce exact sequences from the input ?

Thanks in advance for your input 🙂