no code implementations • insights (ACL) 2022 • Itsuki Okimura, Machel Reid, Makoto Kawano, Yutaka Matsuo
The reason for this is that within NLP, the impact of proposed data augmentation methods on performance has not been evaluated in a unified manner, and effective data augmentation methods are unclear.
1 code implementation • 3 Apr 2024 • Takeshi Kojima, Itsuki Okimura, Yusuke Iwasawa, Hitomi Yanaka, Yutaka Matsuo
Additionally, we tamper with less than 1% of the total neurons in each model during inference and demonstrate that tampering with a few language-specific neurons drastically changes the probability of target language occurrence in text generation.