BankGPT: the use of large language models in official communications 1
Claudia Biancotti, C. Camassa, Marco Fruzzetti et al.
· 0 citations
2 papers indexed here
We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.
Not the right person? Other researchers publish under this name.
#artificial intelligence
May 2026
Do as I Say, Not as I Do: Instruction-Induction Conflict in LLMs
It is suggested that instruction-following remains brittle under induction pressure even for otherwise capable models, and that output diversity, rather than semantic engagement with the input, is the primary factor predicting robustness.
C. Camassa, Derek Shiller
· arXiv.org · 0 citations