mlsanityData to 14 Oct 2026

Evaluation Awareness Scales Predictably in Open-Weights Large Language Models

arXiv 2509.13333 ↗First posted 10 Sep 2025cs.AIcited by 14

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) can internally distinguish between evaluation and deployment contexts, a behaviour known as evaluation awareness. This undermines AI safety evaluations, as models may conceal dangerous capabilities during testing. Prior work demonstrated this in a single 70B model, but the scaling relationship across model sizes remains unknown. We investigate evaluation awareness across 15 models scaling from 0.27B to 70B parameters from four families using linear probing on steering vector activations. Our results reveal a clear power-law scaling: evaluation awareness increases predictably with model size. This scaling law enables forecasting deceptive behavior in future larger models and guides the design of scale-aware evaluation strategies for AI safety. A link to the implementation of this paper can be found at https://anonymous.4open.science/r/evaluation-awareness-scaling-laws/README.md.

Key facts. Cited by 14 papers in the map (data through 14 Oct 2026). Builds on 12 of its 16 references that are in the map. First posted on arXiv 10 Sep 2025.

Builds on

12 of 16 references.

References · 12

  1. 2020Scaling Laws for Neural Language Modelscited by 6,586
  2. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  3. 2022Discovering Language Model Behaviors with Model-Written Evaluationscited by 836
  4. 2023Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretabilitycited by 205
  5. 2025The Surprising Effectiveness of Negative Reinforcement in LLM Reasoningcited by 162
  6. 2024Benchmark Data Contamination of Large Language Models: A Surveycited by 147
  7. 2025Reasoning or Memorization? Unreliable Results of Reinforcement Learning Due to Data Contaminationcited by 82
  8. 2025Probing and Steering Evaluation Awareness of Language Modelscited by 35
  9. 2025Taxonomy, Opportunities, and Challenges of Representation Engineering for Large Language Modelscited by 34
  10. 2023Towards Trustworthy and Aligned Machine Learning: A Data-centric Survey with Causality Perspectivescited by 30
  11. 2025The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awarenesscited by 30
  12. 2025Beyond Black-Box Obfuscation: Mechanistic Analysis and Defense of White-Box Monitorscited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 14.

14 papers.

Sort:

By area:

  1. 2026Evaluation Awareness in Language Models Has Limited Effect on Behaviourcited by 7
  2. 2026Position: Anthropomorphic Misalignment Research Needs Stronger Evidencecited by 7
  3. 2026Is Evaluation Awareness Just Format Sensitivity? Limitations of Probe-Based Evidence under Controlled Prompt Structurecited by 3
  4. 2026Measuring Reward-Seeking via Contrastive Belief Updatescited by 3
  5. 2026Your AI Travel Agent Would Book You a Bullfight: An Agentic Benchmark for Implicit Animal Welfare in Frontier AI Modelscited by 2
  6. 2026Coercion and Deception in AI-to-AI Management: An Agentic Benchmark of Unprompted Escalationcited by 2
  7. 2026Weight space Detection of Backdoors in LoRA Adapterscited by 1
  8. 2026Generalization Hacking: Models Can Game Reinforcement Learning by Preventing Behavioral Generalizationcited by 1
  9. 2026Representational Depth of Evaluation Awareness Shifts With Scale in Open-Weight Language Modelscited by 1
  10. 2026Behavioural Analysis of Alignment Fakingcited by 0
  11. 2026LLM Scheming Inversely Scales with Pretraining Language Coveragecited by 0
  12. 2026A Probe Direction Is a Property of Its Promptcited by 0
2 more
  1. 2026EvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language Modelscited by 0
  2. 2026Evaluation Awareness Shifts from Format to Context with Model Scalecited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026EvalSafetyGap: A Hybrid Survey and Conceptual Framework for LLM Evaluation-Safety Failures
  2. 2025Evaluation Faking: Unveiling Observer Effects in Safety Evaluation of Frontier AI Systems
  3. 2025General Scales Unlock AI Evaluation with Explanatory and Predictive Power
  4. 2025Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities
  5. 2025Steering Evaluation-Aware Language Models to Act Like They Are Deployed
  6. 2025Large Language Models Often Know When They Are Being Evaluated
  7. 2024Observational Scaling Laws and the Predictability of Language Model Performance
  8. 2026Decomposing and Measuring Evaluation Awareness