Tailoring self-rationalizers with multi-reward distillation

S Ramnath, B Joshi, S Hallinan, X Lu, LH Li… - arXiv preprint arXiv …, 2023 - arxiv.org
Large language models (LMs) are capable of generating free-text rationales to aid question
answering. However, prior work 1) suggests that useful self-rationalization is emergent only …

D-separation for causal self-explanation

W Liu, J Wang, H Wang, R Li, Z Deng… - Advances in Neural …, 2024 - proceedings.neurips.cc
Rationalization aims to strengthen the interpretability of NLP models by extracting a subset
of human-intelligible pieces of their inputting texts. Conventional works generally employ the …

MGR: multi-generator based rationalization

W Liu, H Wang, J Wang, R Li, X Li, Y Zhang… - arXiv preprint arXiv …, 2023 - arxiv.org
Rationalization is to employ a generator and a predictor to construct a self-explaining NLP
model in which the generator selects a subset of human-intelligible pieces of the input text to …

CAVE: Controllable Authorship Verification Explanations

S Ramnath, K Pandey, E Boschee, X Ren - arXiv preprint arXiv …, 2024 - arxiv.org
Authorship Verification (AV)(do two documents have the same author?) is essential in many
sensitive real-life applications. AV is often used in proprietary domains that require a private …

AGR: Reinforced causal agent-guided self-explaining rationalization

Y Zhao, Z Wang, X Li, J Liang, R Li - Proceedings of the 62nd …, 2024 - aclanthology.org
Most existing rationalization approaches are susceptible to degeneration accumulation due
to a lack of effective control over the learning direction of the model during training. To …

Mare: Multi-aspect rationale extractor on unsupervised rationale extraction

H Jiang, J Duan, Z Qu, J Wang - arXiv preprint arXiv:2410.03531, 2024 - arxiv.org
Unsupervised rationale extraction aims to extract text snippets to support model predictions
without explicit rationale annotation. Researchers have made many efforts to solve this task …

Pde+: Enhancing generalization via pde with adaptive distributional diffusion

Y Yuan, B Xu, B Lin, L Hou, F Sun, H Shen… - Proceedings of the AAAI …, 2024 - ojs.aaai.org
The generalization of neural networks is a central challenge in machine learning, especially
concerning the performance under distributions that differ from training ones. Current …

Interventional rationalization

L Yue, Q Liu, L Wang, Y An, Y Du… - Proceedings of the 2023 …, 2023 - aclanthology.org
Selective rationalizations improve the explainability of neural networks by selecting a
subsequence of the input (ie, rationales) to explain the prediction results. Although existing …

Is the MMI Criterion Necessary for Interpretability? Degenerating Non-causal Features to Plain Noise for Self-Rationalization

W Liu, Z Deng, Z Niu, J Wang, H Wang… - arXiv preprint arXiv …, 2024 - arxiv.org
An important line of research in the field of explainability is to extract a small subset of crucial
rationales from the full input. The most widely used criterion for rationale extraction is the …

Interlocking-free Selective Rationalization Through Genetic-based Learning

F Ruggeri, G Signorelli - arXiv preprint arXiv:2412.10312, 2024 - arxiv.org
A popular end-to-end architecture for selective rationalization is the select-then-predict
pipeline, comprising a generator to extract highlights fed to a predictor. Such a cooperative …