Hide and Seek with LLMs: An Adversarial Game for Sneaky Error Generation and Self-Improving Diagnosis

• URL: http://arxiv.org/abs/2508.03396v1
• Date: Tue, 05 Aug 2025 12:45:21 GMT
• Title: Hide and Seek with LLMs: An Adversarial Game for Sneaky Error Generation and Self-Improving Diagnosis
• Authors: Rui Zou, Mengqi Wei, Yutao Zhu, Jirong Wen, Xin Zhao, Jing Chen,
• Abstract summary: We propose Hide and Seek Game (HSG), a dynamic adversarial framework for error generation and diagnosis.<n>HSG involves two adversarial roles: Sneaky, which "hides" by generating subtle, deceptive reasoning errors, and Diagnosis, which "seeks" to accurately detect them.<n> Experiments on several math reasoning tasks show that HSG significantly boosts error diagnosis, achieving 16.8%--31.4% higher accuracy than baselines like GPT-4o.
• Score: 51.88592148135258
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) excel in reasoning and generation across domains, but still struggle with identifying and diagnosing complex errors. This stems mainly from training objectives that prioritize correct answers, limiting exposure to and learning from errors. While recent studies have begun to address this by introducing error signals, most rely on shallow, static errors, restricting improvement in deep diagnostic ability. To overcome this, we propose Hide and Seek Game (HSG), a dynamic adversarial framework for error generation and diagnosis, and evaluate it on mathematical problem-solving. HSG involves two adversarial roles: Sneaky, which "hides" by generating subtle, deceptive reasoning errors, and Diagnosis, which "seeks" to accurately detect them. Through adversarial co-evolution, both error stealth and diagnostic precision are enhanced. Experiments on several math reasoning tasks show that HSG significantly boosts error diagnosis, achieving 16.8\%--31.4\% higher accuracy than baselines like GPT-4o. We also release a challenging dataset of deceptive errors and diagnostic annotations as a benchmark for future research.

Related papers

• Text-Guided Multi-Instance Learning for Scoliosis Screening via Gait Video Analysis [33.88520129574637]
  Early-stage scoliosis is difficult to detect, particularly in adolescents, where delayed diagnosis can lead to serious health issues.<n>Traditional X-ray-based methods carry radiation risks and rely heavily on clinical expertise, limiting their use in large-scale screenings.<n>We propose a Text-Guided Multi-Instance Learning Network (TG-MILNet) for non-invasive scoliosis detection using gait videos.
  arXiv  Detail & Related papers  (2025-07-01T22:13:27Z)
• Not All Errors Are Equal: Investigation of Speech Recognition Errors in Alzheimer's Disease Detection [62.942077348224046]
  Speech recognition plays an important role in automatic detection of Alzheimer's disease (AD)<n>Recent studies have revealed a non-linear relationship between word error rates (WER) and AD detection performance.<n>This work presents a series of analyses to explore the effect of ASR transcription errors in BERT-based AD detection systems.
  arXiv  Detail & Related papers  (2024-12-09T09:32:20Z)
• Subtle Errors in Reasoning: Preference Learning via Error-injected Self-editing [59.405145971637204]
  We propose a novel preference learning framework called eRror-Injected Self-Editing (RISE)<n>RISE injects predefined subtle errors into pivotal tokens in reasoning or steps to construct hard pairs for error mitigation.<n>Experiments validate the effectiveness of RISE, with preference learning on Qwen2-7B-Instruct yielding notable improvements of 3.0% on GSM8K and 7.9% on MATH with only 4.5K training samples.
  arXiv  Detail & Related papers  (2024-10-09T07:43:38Z)
• On the Within-class Variation Issue in Alzheimer's Disease Detection [60.08015780474457]
  Alzheimer's Disease (AD) detection employs machine learning classification models to distinguish between individuals with AD and those without.<n>In this work, we found using a sample score estimator can generate sample-specific soft scores aligning with cognitive scores.<n>We propose two simple yet effective methods: Soft Target Distillation (SoTD) and Instance-level Re-balancing (InRe)
  arXiv  Detail & Related papers  (2024-09-22T02:06:05Z)
• Towards Reducing Diagnostic Errors with Interpretable Risk Prediction [18.474645862061426]
  We propose a method to use LLMs to identify pieces of evidence in patient EHR data that indicate increased or decreased risk of specific diagnoses. Our ultimate aim is to increase access to evidence and reduce diagnostic errors.
  arXiv  Detail & Related papers  (2024-02-15T17:05:48Z)
• DDxT: Deep Generative Transformer Models for Differential Diagnosis [51.25660111437394]
  We show that a generative approach trained with simpler supervised and self-supervised learning signals can achieve superior results on the current benchmark. The proposed Transformer-based generative network, named DDxT, autoregressively produces a set of possible pathologies, i.e., DDx, and predicts the actual pathology using a neural network.
  arXiv  Detail & Related papers  (2023-12-02T22:57:25Z)
• PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation [7.508437260320598]
  We propose diagnosis-driven prompts for medical report generation (PromptMRG) PromptMRG is based on encoder-decoder architecture with an extra disease classification branch. Cross-modal feature enhancement retrieves similar reports from the database to assist the diagnosis of a query image.
  arXiv  Detail & Related papers  (2023-08-24T07:10:31Z)
• Deep Reinforcement Learning Framework for Thoracic Diseases Classification via Prior Knowledge Guidance [49.87607548975686]
  The scarcity of labeled data for related diseases poses a huge challenge to an accurate diagnosis. We propose a novel deep reinforcement learning framework, which introduces prior knowledge to direct the learning of diagnostic agents. Our approach's performance was demonstrated using the well-known NIHX-ray 14 and CheXpert datasets.
  arXiv  Detail & Related papers  (2023-06-02T01:46:31Z)
• Assessing glaucoma in retinal fundus photographs using Deep Feature Consistent Variational Autoencoders [63.391402501241195]
  glaucoma is challenging to detect since it remains asymptomatic until the symptoms are severe. Early identification of glaucoma is generally made based on functional, structural, and clinical assessments. Deep learning methods have partially solved this dilemma by bypassing the marker identification stage and analyzing high-level information directly to classify the data.
  arXiv  Detail & Related papers  (2021-10-04T16:06:49Z)
• Learning from Subjective Ratings Using Auto-Decoded Deep Latent Embeddings [23.777855250882244]
  Managing subjectivity in labels is a fundamental problem in medical imaging analysis. We introduce auto-decoded deep latent embeddings (ADDLE) ADDLE explicitly models the tendencies of each rater using an auto-decoder framework.
  arXiv  Detail & Related papers  (2021-04-12T15:40:42Z)

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.