Responsibilities‍

The intern will lead a literature review across linguistics and the philosophy of language to develop a taxonomy of grounding and support relations for AI‑generated statements. They will contribute definitions, examples, and decision rules that make the taxonomy operational for both human annotators and LLM‑as‑judge evaluators.

The intern will design a benchmark: selecting suitable source corpora (including recent groundedness datasets), constructing statement–source pairs, and writing clear annotation guidelines. They will run a human annotation study, potentially crowdsourcing. Where applicable, they will help prepare bespoke annotation tooling.

The intern will evaluate frontier models’ ability to classify grounding categories and compare LLM‑as‑judge performance to human raters. They will co‑author an academic paper describing the taxonomy, dataset, and findings.

‍

‍

‍Qualifications‍

Required/Minimum Qualifications:  

• Is currently enrolled in or has recently completed a PhD program in human-computer interaction, natural language processing, linguistics, philosophy of language.

• Experience in coding for research prototypes, data processing, and simple statistical analysis.

• Is comfortable reviewing and engaging with interdisciplinary literature.

• High fluency in spoken and written English.

‍

‍

Preferred/Additional Qualifications:

• Experience developing AI benchmarks.

• Experience in running human annotation or crowdsourcing studies.

• Experience building prototypes that leverage language model APIs.

• Experience in prompt design and engineering.

• Research publications at top conferences and journals in HCI/ML/NLP.

‍