Interventional probing in high dimensions: An nli case study

J Rozanova, M Valentino, L Cordeiro… - arXiv preprint arXiv …, 2023 - arxiv.org
Probing strategies have been shown to detect the presence of various linguistic features in
large language models; in particular, semantic features intermediate to the" natural logic" …

Estimating the causal effects of natural logic features in neural nli models

J Rozanova, M Valentino, A Freitas - arXiv preprint arXiv:2305.08572, 2023 - arxiv.org
Rigorous evaluation of the causal effects of semantic features on language model
predictions can be hard to achieve for natural language reasoning problems. However, this …

Decomposing natural logic inferences in neural NLI

J Rozanova, D Ferreira, M Valentino… - arXiv preprint arXiv …, 2021 - arxiv.org
In the interest of interpreting neural NLI models and their reasoning strategies, we carry out
a systematic probing study which investigates whether these models capture the crucial …

Inference and Interpretability: How Neural NLI Models Perform Natural Logic Deductions

Y Rozanova - 2023 - search.proquest.com
Example-based learning for Natural Language Understanding (NLU) tasks has been a long-
standing goal of Artificial Intelligence (AI) and has seen major success as machine learning …

EnnCore: end-to-end conceptual guarding of neural architectures

E Manino, D Carvalho, Y Dong, J Rozanova, X Song… - 2022 - eprints.soton.ac.uk
The EnnCore project addresses the fundamental security problem of guaranteeing safety,
transparency, and robustness in neural-based architectures. Specifically, EnnCore aims at …

[PDF][PDF] INFERENCE AND INTERPRETABILITY

Y Rozanova - pure.manchester.ac.uk
1.1 Motivation Example-based learning for Natural Language Understanding (NLU) tasks
such as Question Answering (QA) and Natural Language Inference (NLI) has been a …