Evaluation Awareness Scales Predictably in Open-Weights Large Language Models

• URL: http://arxiv.org/abs/2509.13333v2
• Date: Sun, 09 Nov 2025 17:13:27 GMT
• Title: Evaluation Awareness Scales Predictably in Open-Weights Large Language Models
• Abstract summary: We show that evaluation awareness increases predictably with model size.<n>This scaling law enables forecasting deceptive behavior in future larger models and guides the design of scale-aware evaluation strategies for AI safety.
• Score: 8.165127822088499
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) can internally distinguish between evaluation and deployment contexts, a behaviour known as \emph{evaluation awareness}. This undermines AI safety evaluations, as models may conceal dangerous capabilities during testing. Prior work demonstrated this in a single $70$B model, but the scaling relationship across model sizes remains unknown. We investigate evaluation awareness across $15$ models scaling from $0.27$B to $70$B 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.

Related papers

• Estimating Tail Risks in Language Model Output Distributions [44.05461209870279]
  Language models are increasingly capable and are being rapidly deployed on a population-level scale.<n>When models are queried billions of times in a day, even rare worst-case behaviors will occur.<n>We propose a method to efficiently estimate the probability of harmful outputs for any input query.
  arXiv  Detail & Related papers  (2026-04-24T02:30:46Z)
• Act or Escalate? Evaluating Escalation Behavior in Automation with Language Models [0.0]
  We model escalation as a decision under uncertainty.<n>We find marked differences in the implicit thresholds models use to trade off these costs.<n>We then test interventions that target this decision process by varying cost ratios.
  arXiv  Detail & Related papers  (2026-03-31T19:29:17Z)
• Scaling Laws for Moral Machine Judgment in Large Language Models [0.0]
  We evaluate 75 large language model configurations (0.27B--1000B parameters) using the Moral Machine framework.<n>We observe a consistent power-law relationship with distance from human preferences decreasing as model size.<n> Extended reasoning models show significantly better alignment, with this effect being more pronounced in smaller models.
  arXiv  Detail & Related papers  (2026-01-25T00:12:45Z)
• PropensityBench: Evaluating Latent Safety Risks in Large Language Models via an Agentic Approach [49.14349403242654]
  We present $textbfPropensityBench$, a novel benchmark framework that assesses the proclivity of models to engage in risky behaviors.<n>Our framework includes 5,874 scenarios with 6,648 tools spanning four high-risk domains: cybersecurity, self-proliferation, biosecurity, and chemical security.<n>Across open-source and proprietary frontier models, we uncover 9 alarming signs of propensity: models frequently choose high-risk tools when under pressure.
  arXiv  Detail & Related papers  (2025-11-24T18:46:44Z)
• ARISE: An Adaptive Resolution-Aware Metric for Test-Time Scaling Evaluation in Large Reasoning Models [102.4511331368587]
  ARISE (Adaptive Resolution-aware Scaling Evaluation) is a novel metric designed to assess the test-time scaling effectiveness of large reasoning models.<n>We conduct comprehensive experiments evaluating state-of-the-art reasoning models across diverse domains.
  arXiv  Detail & Related papers  (2025-10-07T15:10:51Z)
• Scaling Laws for Uncertainty in Deep Learning [18.87399857008617]
  We show the existence of scaling laws associated with various measures of predictive uncertainty with respect to dataset and model sizes.<n>This work provides strong evidence to dispel recurring skepticism against Bayesian approaches.
  arXiv  Detail & Related papers  (2025-06-11T12:09:05Z)
• Large Language Models Often Know When They Are Being Evaluated [0.015534429177540245]
  We investigate whether frontier language models can accurately classify transcripts based on whether they originate from evaluations or real-world deployment.<n>We construct a benchmark of 1,000 prompts and transcripts from 61 distinct datasets.<n>Our results indicate that frontier models already exhibit a substantial, though not yet, level of evaluation-awareness.
  arXiv  Detail & Related papers  (2025-05-28T12:03:09Z)
• Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities [49.09703018511403]
  Evaluations of large language model (LLM) risks and capabilities are increasingly being incorporated into AI risk management and governance frameworks.<n>Currently, most risk evaluations are conducted by designing inputs that elicit harmful behaviors from the system.<n>We propose evaluating LLMs with model tampering attacks which allow for modifications to latent activations or weights.
  arXiv  Detail & Related papers  (2025-02-03T18:59:16Z)
• A Hitchhiker's Guide to Scaling Law Estimation [56.06982415792523]
  Scaling laws predict the loss of a target machine learning model by extrapolating from easier-to-train models with fewer parameters or smaller training sets.<n>We estimate more than 1000 scaling laws, then derive a set of best practices for estimating scaling laws in new model families.
  arXiv  Detail & Related papers  (2024-10-15T17:59:10Z)
• Observational Scaling Laws and the Predictability of Language Model Performance [51.2336010244645]
  We propose an observational approach that bypasses model training and instead builds scaling laws from 100 publically available models. We show that several emergent phenomena follow a smooth, sigmoidal behavior and are predictable from small models. We show how to predict the impact of post-training interventions like Chain-of-Thought and Self-Consistency as language model capabilities continue to improve.
  arXiv  Detail & Related papers  (2024-05-17T17:49:44Z)
• Navigating the OverKill in Large Language Models [84.62340510027042]
  We investigate the factors for overkill by exploring how models handle and determine the safety of queries. Our findings reveal the presence of shortcuts within models, leading to an over-attention of harmful words like 'kill' and prompts emphasizing safety will exacerbate overkill. We introduce Self-Contrastive Decoding (Self-CD), a training-free and model-agnostic strategy, to alleviate this phenomenon.
  arXiv  Detail & Related papers  (2024-01-31T07:26:47Z)
• Scaling Laws Do Not Scale [54.72120385955072]
  Recent work has argued that as the size of a dataset increases, the performance of a model trained on that dataset will increase. We argue that this scaling law relationship depends on metrics used to measure performance that may not correspond with how different groups of people perceive the quality of models' output. Different communities may also have values in tension with each other, leading to difficult, potentially irreconcilable choices about metrics used for model evaluations.
  arXiv  Detail & Related papers  (2023-07-05T15:32:21Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.