Papers by Elizabeth Daly
Language Models in Dialogue: Conversational Maxims for Human-AI Interactions (2024.findings-emnlp)
Copied to clipboard
Erik Miehling, Manish Nagireddy, Prasanna Sattigeri, Elizabeth Daly, David Piorkowski, John Richards
| Challenge: | Modern language models exhibit some inherent shortcomings, particularly in conversational settings. |
| Approach: | They propose a set of maxims for describing effective human-AI conversation that include quantity, quality, relevance, manner, benevolence, and transparency. |
| Outcome: | The proposed maxims are applied to human-AI interactions and are based on extensive research from the social science and AI communities. |
Know Who Your Friends Are: Understanding Social Connections from Unstructured Text (N18-5)
Copied to clipboard
Léa Deleris, Francesca Bonin, Elizabeth Daly, Stéphane Deparis, Yufang Hou, Charles Jochim, Yassine Lassoued, Killian Levacher
| Challenge: | Having an understanding of interpersonal relationships is helpful in many contexts. |
| Approach: | They propose a system that extracts qualitative and quantitative information from texts and aggregates it to provide a condensed view of relationships. |
| Outcome: | The proposed system extracts qualitative and quantitative information elements about interactions and aggregates those to provide a condensed view of relationships. |
Ranking Large Language Models without Ground Truth (2024.findings-acl)
Copied to clipboard
| Challenge: | Evaluation and ranking of large language models has become a problem with the proliferation of these models and their impact. |
| Approach: | They propose to rank large language models without access to ground truth or reference responses . they propose to use triplets of models to evaluate the other two, correctly identifying the worst model in the triplet with high probability. |
| Outcome: | The proposed method reliably recovers true rankings without reference data on generative tasks. |