Abstract
One strength of modern language models is their ability to incorporate information from a user-input context when answering queries. However, they are not equally sensitive to the subtle changes to that context.To quantify this, Du et al. (2024) gives an information-theoretic metric to measure such sensitivity. Their metric, susceptibility, is defined as the degree to which contexts can influence a model’s response to a query at a distributional level.However, exactly computing susceptibility is difficult and, thus, Du et al. (2024) falls back on a Monte Carlo approximation.Due to the large number of samples required, the Monte Carlo approximation is inefficient in practice. As a faster alternative, we propose Fisher susceptibility, an efficient method to estimate the susceptibility based on Fisher information.Empirically, we validate that Fisher susceptibility is comparable to Monte Carlo estimated susceptibility across a diverse set of query domains despite its being 70× faster.Exploiting the improved efficiency, we apply Fisher susceptibility to analyze factors affecting the susceptibility of language models.We observe that larger models are as susceptible as smaller ones.- Anthology ID:
- 2024.findings-emnlp.386
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2024
- Month:
- November
- Year:
- 2024
- Address:
- Miami, Florida, USA
- Editors:
- Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 6615–6626
- Language:
- URL:
- https://aclanthology.org/2024.findings-emnlp.386
- DOI:
- 10.18653/v1/2024.findings-emnlp.386
- Cite (ACL):
- Tianyu Liu, Kevin Du, Mrinmaya Sachan, and Ryan Cotterell. 2024. Efficiently Computing Susceptibility to Context in Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 6615–6626, Miami, Florida, USA. Association for Computational Linguistics.
- Cite (Informal):
- Efficiently Computing Susceptibility to Context in Language Models (Liu et al., Findings 2024)
- PDF:
- https://preview.aclanthology.org/dois-2013-emnlp/2024.findings-emnlp.386.pdf