What type of study is this?

This is a Quantitative Study study.

September 30, 2025Open Access

How to Improve the Robustness of Closed-Source Models on NLI

View Full Paper

Ask AI

Bookmark

Authors

JSJoe StaceyNational Hospital for Neurology and Neurosurgery LALisa Alazraki AUAran Ubhi

Discussion

Member takes

Implication

This work investigates data-centric strategies to improve robustness in closed-source large language models, suggesting implications for out-of-distribution performance.

Key Points

Improving robustness in closed-source large language models can increase performance on out-of-distribution data.
Upsampling challenging training examples enhances robustness by up to 1.5% for complex out-of-distribution datasets.
Replacing training data with LLM-generated examples boosts robustness by 3.7% for less complex out-of-distribution datasets.
Closed-source autoregressive LLMs outperform commonly used encoder models in robustness, indicating a need for adjusted baselines.

Cite This Study

Stacey et al. (2025) studied this question.

synapsesocial.com/papers/68dc12cc8a7d58c25ebb0afe https://doi.org/10.48550/arxiv.2505.20209

View Full Paper

Ask AI

Bookmark

Implication

This work investigates data-centric strategies to improve robustness in closed-source large language models, suggesting implications for out-of-distribution performance.

How to Improve the Robustness of Closed-Source Models on NLI

Discussion

Member takes

Implication

Key Points

Cite This Study

How to Improve the Robustness of Closed-Source Models on NLI

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider