[2603.19426] Is Evaluation Awareness Just Format Sensitivity? Limitations of Probe-Based Evidence under Controlled Prompt Structure

Computer Science > Computation and Language arXiv:2603.19426 (cs) [Submitted on 19 Mar 2026] Title:Is Evaluation Awareness Just Format Sensitivity? Limitations of Probe-Based Evidence under Controlled Prompt Structure Authors:Viliana Devbunova View a PDF of the paper titled Is Evaluation Awareness Just Format Sensitivity? Limitations of Probe-Based Evidence under Controlled Prompt Structure, by Viliana Devbunova View PDF HTML (experimental) Abstract:Prior work uses linear probes on benchmark prompts as evidence of evaluation awareness in large language models. Because evaluation context is typically entangled with benchmark format and genre, it is unclear whether probe-based signals reflect context or surface structure. We test whether these signals persist under partial control of prompt format using a controlled 2x2 dataset and diagnostic rewrites. We find that probes primarily track benchmark-canonical structure and fail to generalize to free-form prompts independent of linguistic style. Thus, standard probe-based methodologies do not reliably disentangle evaluation context from structural artifacts, limiting the evidential strength of existing results. Comments: Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI) ACM classes: I.2.7 Cite as: arXiv:2603.19426 [cs.CL]   (or arXiv:2603.19426v1 [cs.CL] for this version)   https://doi.org/10.48550/arXiv.2603.19426 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission hist...