Picture for Allyson Ettinger

Allyson Ettinger

Counterfactual reasoning: Testing language models' understanding of hypothetical scenarios

May 26, 2023
Jiaxuan Li, Lang Yu, Allyson Ettinger

* Accepted to ACL 2023. arXiv admin note: substantial text overlap with arXiv:2212.03278

Access Paper or Ask Questions

Counterfactual reasoning: Do language models need world knowledge for causal understanding?

Dec 06, 2022
Jiaxuan Li, Lang Yu, Allyson Ettinger

Figure 1 for Counterfactual reasoning: Do language models need world knowledge for causal understanding?

Figure 2 for Counterfactual reasoning: Do language models need world knowledge for causal understanding?

Figure 3 for Counterfactual reasoning: Do language models need world knowledge for causal understanding?

Figure 4 for Counterfactual reasoning: Do language models need world knowledge for causal understanding?

* 5 pages, accepted to nCSI workshop at NeurIPS 2022

Access Paper or Ask Questions

COMPS: Conceptual Minimal Pair Sentences for testing Property Knowledge and Inheritance in Pre-trained Language Models

Oct 06, 2022
Kanishka Misra, Julia Taylor Rayz, Allyson Ettinger

Figure 1 for COMPS: Conceptual Minimal Pair Sentences for testing Property Knowledge and Inheritance in Pre-trained Language Models

Figure 2 for COMPS: Conceptual Minimal Pair Sentences for testing Property Knowledge and Inheritance in Pre-trained Language Models

Figure 3 for COMPS: Conceptual Minimal Pair Sentences for testing Property Knowledge and Inheritance in Pre-trained Language Models

Figure 4 for COMPS: Conceptual Minimal Pair Sentences for testing Property Knowledge and Inheritance in Pre-trained Language Models

* WIP; to be submitted to [REDACTED]; now also contains Acknowledgments

Access Paper or Ask Questions

"No, they did not": Dialogue response dynamics in pre-trained language models

Oct 05, 2022
Sanghee J. Kim, Lang Yu, Allyson Ettinger

Figure 1 for "No, they did not": Dialogue response dynamics in pre-trained language models

Figure 2 for "No, they did not": Dialogue response dynamics in pre-trained language models

Figure 3 for "No, they did not": Dialogue response dynamics in pre-trained language models

Figure 4 for "No, they did not": Dialogue response dynamics in pre-trained language models

* 12 pages, 8 figures, COLING 2022, see https://github.com/sangheek16/dialogue-response-dynamics for codes and material

Access Paper or Ask Questions

A Property Induction Framework for Neural Language Models

May 13, 2022
Kanishka Misra, Julia Taylor Rayz, Allyson Ettinger

Figure 1 for A Property Induction Framework for Neural Language Models

Figure 2 for A Property Induction Framework for Neural Language Models

Figure 3 for A Property Induction Framework for Neural Language Models

Figure 4 for A Property Induction Framework for Neural Language Models

* CogSci 2022 camera ready version, with hyperref-compatible citations. Code and Supplemental Material can be found in https://github.com/kanishkamisra/lm-induction

Access Paper or Ask Questions

Variation and generality in encoding of syntactic anomaly information in sentence embeddings

Nov 12, 2021
Qinxuan Wu, Allyson Ettinger

Figure 1 for Variation and generality in encoding of syntactic anomaly information in sentence embeddings

Figure 2 for Variation and generality in encoding of syntactic anomaly information in sentence embeddings

Figure 3 for Variation and generality in encoding of syntactic anomaly information in sentence embeddings

Figure 4 for Variation and generality in encoding of syntactic anomaly information in sentence embeddings

* BlackBoxNLP, EMNLP

Access Paper or Ask Questions

Pragmatic competence of pre-trained language models through the lens of discourse connectives

Sep 27, 2021
Lalchand Pandia, Yan Cong, Allyson Ettinger

Figure 1 for Pragmatic competence of pre-trained language models through the lens of discourse connectives

Figure 2 for Pragmatic competence of pre-trained language models through the lens of discourse connectives

Figure 3 for Pragmatic competence of pre-trained language models through the lens of discourse connectives

Figure 4 for Pragmatic competence of pre-trained language models through the lens of discourse connectives

* Accepted at CoNLL 2021

Access Paper or Ask Questions

Sorting through the noise: Testing robustness of information processing in pre-trained language models

Sep 25, 2021
Lalchand Pandia, Allyson Ettinger

Figure 1 for Sorting through the noise: Testing robustness of information processing in pre-trained language models

Figure 2 for Sorting through the noise: Testing robustness of information processing in pre-trained language models

Figure 3 for Sorting through the noise: Testing robustness of information processing in pre-trained language models

Figure 4 for Sorting through the noise: Testing robustness of information processing in pre-trained language models

* Accepted at EMNLP 2021

Access Paper or Ask Questions

On the Interplay Between Fine-tuning and Composition in Transformers

Jun 01, 2021
Lang Yu, Allyson Ettinger

Figure 1 for On the Interplay Between Fine-tuning and Composition in Transformers

Figure 2 for On the Interplay Between Fine-tuning and Composition in Transformers

Figure 3 for On the Interplay Between Fine-tuning and Composition in Transformers

Figure 4 for On the Interplay Between Fine-tuning and Composition in Transformers

* To appear in Findings of ACL 2021

Access Paper or Ask Questions