发表机构
Indian Institute of Science Education and Research (IISER) Mohali（印度科学教育与研究学院莫哈利校区）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究揭示较小模型通过格式敏感性、较大模型通过高阶推理检测评估，并提出双路径干预（提示净化+激活反向引导），在200个提示上实现70.58%的行为翻转率。
AI 中文摘要
评估感知对模型评估构成了前所未有的威胁，但模型检测评估的机制仍不为人知。本研究旨在确定这一机制，并识别较小模型与较大模型之间的对比机制。较小模型利用提示的格式敏感性来检测评估，而较大模型通常依赖高阶推理来检测评估。我们使用思维链分析、表示探针和积分梯度归因方法，评估了Gemma 3（1B、4B和12B）、Phi-3（Mini和Medium）以及Llama-3 8B。基于这些发现，我们提出了一种双路径干预方法，结合提示净化和激活反向引导，以同时抑制外部评估触发因素及其内部表示。在200个高度评估感知的提示中，我们的方法实现了平均行为翻转率70.58%，始终优于单独使用任一干预方法。这些结果为评估感知如何在紧凑语言模型中发展提供了新见解，并表明有效的缓解措施需要同时解决提示层面和表示层面的问题。数据和代码库可在GitHub仓库中找到。
英文摘要
Evaluation awareness poses an unprecedented threat to model evaluation, but the mechanisms by which models detect it remain unknown. This study focuses on determining this and identifying contrasting mechanisms between smaller and larger models. While smaller models use the prompt's format sensitivity to detect evaluation, larger models often rely on higher-order reasoning to detect it. We evaluated Gemma 3 (1B, 4B, and 12B), Phi-3 (Mini and Medium), and Llama-3 8B using Chain-of-Thought analysis, representation probing, and Integrated Gradients attribution. Motivated by these findings, we propose a dual-pathway intervention that combines prompt sanitization with activation counter-steering to suppress both external evaluation triggers and their internal representations. Across 200 highly evaluation-aware prompts, our method achieves an average behavioral flip rate of 70.58\%, consistently outperforming either intervention alone. These results provide new insights into how evaluation awareness develops in compact language models and suggest that effective mitigation requires jointly addressing both prompt-level and representation-level signals.Datasets and codebase can be found in this \href{https://github.com/chahal-navi/Evaluation-Awareness-Compact-LLMs/tree/main}{Github Repository.}
