SimInstruct : A Responsible Tool for Collecting Scaffolding Dialogues Between Experts and LLM-Simulated Novices
Si Chen, Izzy Molnar, Ting Hua, Peiyu Li, Le Huy Khiem, G. Alex Ambrose, Jim Lang, Ronald Metoyer, Nitesh V. Chawla · 2025
Read SimInstruct : A Responsible Tool for Collecting Scaffolding Dialogues Between Experts and LLM-Simulated Novices by Si Chen, Izzy Molnar, Ting Hua, Peiyu Li, Le Huy Khiem, G. Alex Ambrose, Jim Lang, Ronald Metoyer, Nitesh V. Chawla (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-04428.txt.
Description
Si Chen and co-authors build a tool that fakes novice teachers for coaching data in a 2025 paper. Expert novice talk is gold for training teaching AI, but real novices are scarce and shy to be taped. SimInstruct pairs human experts with language model novices that vary in teaching snags and persona traits.
Experts gave multi turn feedback to the fake novices, building rich scaffolding talks with no real novice at risk. Persona traits like outward or inward bent changed how experts coached. The talks matched real mentoring on depth, and experts found the work engaging. A tuned model beat a top rival on teaching grade.
Catalog record
| CREATOR | Si Chen, Izzy Molnar, Ting Hua, Peiyu Li, Le Huy Khiem, G. Alex Ambrose, Jim Lang, Ronald Metoyer, Nitesh V. Chawla |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-04428