Scaling Trends for Lie Detector Oversight in Preference Learning

• URL: http://arxiv.org/abs/2607.01567v1
• Date: Thu, 02 Jul 2026 00:53:33 GMT
• Title: Scaling Trends for Lie Detector Oversight in Preference Learning
• Abstract summary: In this paper, we scale SOLiD to larger models and evaluate it in more diverse and realistic preference-learning settings.<n>We find favorable scaling: undetected deception drops from 34% for 1B- parameter models to 14% for 405B- parameter models at a detector true positive rate of 99%.<n>However, SOLiD is sensitive to distribution shift between detector training and preference-training data, which can drive detector false positive rates to impractical levels.
• Score: 10.250952047645457
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Deceptive behavior in LLMs is costly to monitor and prevent, motivating approaches such as Scalable Oversight via Lie Detectors (SOLiD) (Cundy & Gleave, 2025), which uses lie detectors to identify responses for review by high-cost labelers. In this paper, we scale SOLiD to larger models and evaluate it in more diverse and realistic preference-learning settings. We find favorable scaling: undetected deception drops from 34% for 1B-parameter models to 14% for 405B-parameter models at a detector true positive rate of 99%, and expensive human labelers can be removed entirely from the fine-tuning phase without a statistically significant increase in deception. However, SOLiD is sensitive to distribution shift between detector training and preference-training data, which can drive detector false positive rates to impractical levels.

Related papers

• Cybersecurity Detection Classification with Reasoning-enabled Language Models [50.72675435043546]
  A major issue in Security Operations Centers (SOCs) is alert fatigue.<n>Prior work prompts or fine-tunes large language models (LLMs) to emit a triage label directly, but does not train them to reason about whether a detection is a genuine threat.<n>We train a chain-of-thought (CoT) reasoning-enabled triage classifier on real, human-labeled Windows endpoint detections.
  arXiv  Detail & Related papers  (2026-07-30T16:22:13Z)
• "Did you lie?" Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms [2.3883252067956118]
  We show that existing trained model organisms often fail testbeds where models verifiably believe the opposite of what they say.<n>We address this with 13 reasoning model organisms whose hidden beliefs are verified in chain-of-thought.<n>On prompted lying, across 31 open-weight models spanning 2B to 1T parameters, all four detectors show positive scaling with model capability.
  arXiv  Detail & Related papers  (2026-06-10T19:21:12Z)
• Uncertainty-Aware Exploratory Direct Preference Optimization for Multimodal Large Language Models [53.15468578562038]
  We propose an Uncertainty-aware Exploratory Direct Preference Optimization (UE-DPO) method for Multimodal Large Language Models (MLLMs)<n>We first quantify the uncertainty from the model's failure to ground token predictions in the given image.<n>We then encourage more learning pressure on visually deficient tokens in preferred samples, and alleviate the over-penalization of beneficial knowledge in dispreferred samples.
  arXiv  Detail & Related papers  (2026-05-06T13:08:12Z)
• Semantics-Preserving Evasion of LLM Vulnerability Detectors [14.476903104601154]
  LLM-based vulnerability detectors are increasingly deployed in security-critical code review.<n>We evaluate detection-time integrity under a semantics-preserving threat model.<n>We introduce a metric of joint robustness across different attack methods/carriers.
  arXiv  Detail & Related papers  (2026-01-30T20:54:27Z)
• Preference Learning with Lie Detectors can Induce Honesty or Evasion [6.488157280516656]
  We show that lie detectors can accurately classify deceptive behavior, but they are not typically used in the training pipeline.<n>We examine whether the learned policy is genuinely more honest, or instead learns to fool the lie detector while remaining deceptive.<n>Off-policy algorithms (DPO) consistently lead to deception rates under 25% for realistic TPRs.
  arXiv  Detail & Related papers  (2025-05-20T00:31:53Z)
• Lie Detector: Unified Backdoor Detection via Cross-Examination Framework [68.45399098884364]
  We propose a unified backdoor detection framework in the semi-honest setting.<n>Our method achieves superior detection performance, improving accuracy by 5.4%, 1.6%, and 11.9% over SoTA baselines.<n> Notably, it is the first to effectively detect backdoors in multimodal large language models.
  arXiv  Detail & Related papers  (2025-03-21T06:12:06Z)
• Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-tuned Contrastive Learning Detector would be Better [21.901523394933076]
  We propose a novel fine-tuned detector, Pecola, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Experiments show that Pecola outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets.
  arXiv  Detail & Related papers  (2024-02-01T01:23:07Z)
• Small Effect Sizes in Malware Detection? Make Harder Train/Test Splits! [51.668411293817464]
  Industry practitioners care about small improvements in malware detection accuracy because their models are deployed to hundreds of millions of machines. Academic research is often restrained to public datasets on the order of ten thousand samples. We devise an approach to generate a benchmark of difficulty from a pool of available samples.
  arXiv  Detail & Related papers  (2023-12-25T21:25:55Z)
• Test-Time Adaptation Induces Stronger Accuracy and Agreement-on-the-Line [65.14099135546594]
  Recent test-time adaptation (TTA) methods drastically strengthen the ACL and AGL trends in models, even in shifts where models showed very weak correlations before. Our results show that by combining TTA with AGL-based estimation methods, we can estimate the OOD performance of models with high precision for a broader set of distribution shifts.
  arXiv  Detail & Related papers  (2023-10-07T23:21:25Z)
• Towards Reducing Labeling Cost in Deep Object Detection [61.010693873330446]
  We propose a unified framework for active learning, that considers both the uncertainty and the robustness of the detector. Our method is able to pseudo-label the very confident predictions, suppressing a potential distribution drift.
  arXiv  Detail & Related papers  (2021-06-22T16:53:09Z)
• Robust and Accurate Object Detection via Adversarial Learning [111.36192453882195]
  This work augments the fine-tuning stage for object detectors by exploring adversarial examples. Our approach boosts the performance of state-of-the-art EfficientDets by +1.1 mAP on the object detection benchmark.
  arXiv  Detail & Related papers  (2021-03-23T19:45:26Z)

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.