Hi! I’m Nick, a second-year PhD student at the University of Vienna supervised by Sebastian Schuster. I’m interested in evaluating and improving LLM agents as collaborators, currently focusing on coding agents. My work studies when agents should ask for clarification versus proceed autonomously (see Ask or Assume?), and how they handle users interrupting and correcting them mid-task in real-world developer-agent interactions. I’ve also worked on evaluating agents’ ability to implement AI research (RExBench).
Previously, I received a BA in Linguistics from the University of Cambridge, followed by an MSc in Speech & Language Processing from the University of Edinburgh. During my time at Edinburgh, I did research in emergent communication between neural networks with Hannah Rohde and Henry Conklin.
Get in touch here: nicholas.edwards[æt]univie.ac.at
news
| Sep 25, 2026 | 🎉 Harbor Adapters and Harbor-Index accepted to the NeurIPS 2026 Evaluations & Datasets track! |
|---|---|
| Aug 21, 2026 | 🎉 Ask or Assume? accepted to EMNLP 2026 Findings and the REALM workshop! See you in Budapest! |
| Apr 17, 2026 | RExBench has been integrated into the Harbor framework! See the announcement for more details. |
| Apr 07, 2026 | 🎉 RExBench accepted to the main ACL 2026 conference! Excited to share this work in San Diego! |
| Sep 15, 2025 | Gave an invited talk on RExBench to Ludwig Schmidt’s group at Stanford/UW, together with Yukyung Lee! |
selected publications
- NeurIPS ’26Harbor Adapters and Harbor-Index: Infrastructure and a Curated Meta-Dataset for Large-Scale Agentic EvaluationIn Advances in Neural Information Processing Systems, Evaluations & Datasets Track, 2026
- CogSci ’23Anaphoric structure emerges between neural networksIn 45th Annual Conference of the Cognitive Science Society: Cognition in Context, 2023