A preprint has a strong model write a skill library for a weaker one, and when the teacher profiles the student model first, task success reaches 86.46%, against 81.04% without profiling.

According to the paper's abstract, when the two models swapped skill libraries each did better with the one written for it, a claim given without figures, and the paper tests only three models. A skill library is a set of ready-written procedures an agent can call while working. In profiling, the teacher first builds a learner model of the student, and a skill goes into the library only if it raises the student's success rate. The authors had Claude-Sonnet-5 teach DeepSeek-V4-Flash, averaging four main runs; the untaught student succeeded 75.00% of the time. Against writing skills without regard to who the student is, profiling adds about 5 percentage points. The abstract doesn't give the task set or the number of tasks; the results are the authors' own and no one else has repeated them (arXiv 2611.13144, 2026-11-09). How we checked →

Source and checks

This item ran in the brief of Nov 12, 2026, alongside 16 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.

Cite this item

SecondSource (2026). "A preprint has a strong model write a skill library for a weaker one, and when the teacher profiles the student model first, task success reaches 86.46%, against 81.04% without profiling.." https://secondsource.io/en/research/2026-11-12-2/. Accessed access date.