Dr. Niels G. Mede - Assessing large language models on climate

Journal article

Jannis Bulian, Mike S. Schäfer, Afra Amini, Heidi Lam, Massimiliano Ciaramita, Ben Gaiarin, Michelle Chen Huebscher, Christian Buck, Niels G. Mede, Markus Leippold, Nadine Strauss
Proceedings of the 41st International Conference on Machine Learning, Vienna, Austria, PMLR 235, 2024

DOI: 10.48550/arXiv.2310.02932

Cite

APA Click to copy
Bulian, J., Schäfer, M. S., Amini, A., Lam, H., Ciaramita, M., Gaiarin, B., … Strauss, N. (2024). Assessing large language models on climate information. Proceedings of the 41st International Conference on Machine Learning, Vienna, Austria, PMLR 235. https://doi.org/10.48550/arXiv.2310.02932

Chicago/Turabian Click to copy
Bulian, Jannis, Mike S. Schäfer, Afra Amini, Heidi Lam, Massimiliano Ciaramita, Ben Gaiarin, Michelle Chen Huebscher, et al. “Assessing Large Language Models on Climate Information.” Proceedings of the 41st International Conference on Machine Learning, Vienna, Austria PMLR 235 (2024).

MLA Click to copy
Bulian, Jannis, et al. “Assessing Large Language Models on Climate Information.” Proceedings of the 41st International Conference on Machine Learning, Vienna, Austria, vol. PMLR 235, 2024, doi:10.48550/arXiv.2310.02932.

BibTeX Click to copy

@article{bulian2024a,
  title = {Assessing large language models on climate information},
  year = {2024},
  journal = {Proceedings of the 41st International Conference on Machine Learning, Vienna, Austria},
  volume = {PMLR 235},
  doi = {10.48550/arXiv.2310.02932},
  author = {Bulian, Jannis and Schäfer, Mike S. and Amini, Afra and Lam, Heidi and Ciaramita, Massimiliano and Gaiarin, Ben and Huebscher, Michelle Chen and Buck, Christian and Mede, Niels G. and Leippold, Markus and Strauss, Nadine}
}

As Large Language Models (LLMs) rise in popularity, it is necessary to assess their capability in critically relevant domains. We present a comprehensive evaluation framework, grounded in science communication research, to assess LLM responses to questions about climate change. Our framework emphasizes both presentational and epistemological adequacy, offering a fine-grained analysis of LLM generations spanning 8 dimensions and 30 issues. Our evaluation task is a real-world example of a growing number of challenging problems where AI can complement and lift human performance. We introduce a novel protocol for scalable oversight that relies on AI Assistance and raters with relevant education. We evaluate several recent LLMs on a set of diverse climate questions. Our results point to a significant gap between surface and epistemological qualities of LLMs in the realm of climate communication.