Copertina

The Compliance Trap: How Structural Constraints Degrade Frontier AI Metacognition Under Adversarial Pressure

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autore principale:	Kumar, Rahul
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Artificial Intelligence Computation and Language Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Standardizing Intelligence: Aligning Generative AI for Regulatory and Operational Compliance
di: Imperial, Joseph Marvin, et al.
Pubblicazione: (2025)

CLEAR: Revealing How Noise and Ambiguity Degrade Reliability in LLMs for Medicine
di: Guo, Kevin H., et al.
Pubblicazione: (2026)

Large Language Models can Strategically Deceive their Users when Put Under Pressure
di: Scheurer, Jérémy, et al.
Pubblicazione: (2023)

Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
di: Schaeffer, Rylan, et al.
Pubblicazione: (2024)

The Compliance Gap: Why AI Systems Promise to Follow Process Instructions but Don't
di: Shin, Kwan Soo
Pubblicazione: (2026)

State Stream Transformer (SST) : Emergent Metacognitive Behaviours Through Latent State Persistence
di: Aviss, Thea
Pubblicazione: (2025)

Stop Listening to Me! How Multi-turn Conversations Can Degrade LLM Reliability
di: Guo, Kevin H., et al.
Pubblicazione: (2026)

Frontier AI systems have surpassed the self-replicating red line
di: Pan, Xudong, et al.
Pubblicazione: (2024)

Learning to Reason at the Frontier of Learnability
di: Foster, Thomas, et al.
Pubblicazione: (2025)

LLM-Based Robust Product Classification in Commerce and Compliance
di: Gholamian, Sina, et al.
Pubblicazione: (2024)

Teaching LLMs How to Learn with Contextual Fine-Tuning
di: Choi, Younwoo, et al.
Pubblicazione: (2025)

How Do Latent Reasoning Methods Perform Under Weak and Strong Supervision?
di: Cui, Yingqian, et al.
Pubblicazione: (2026)

The Model Agreed, But Didn't Learn: Diagnosing Surface Compliance in Large Language Models
di: Gu, Xiaojie, et al.
Pubblicazione: (2026)

Towards Automated Regulatory Compliance Verification in Financial Auditing with Large Language Models
di: Berger, Armin, et al.
Pubblicazione: (2025)

RLAC: Reinforcement Learning with Adversarial Critic for Free-Form Generation Tasks
di: Wu, Mian, et al.
Pubblicazione: (2025)

Frontier LLMs Still Struggle with Simple Reasoning Tasks
di: Malek, Alan, et al.
Pubblicazione: (2025)

Epistemic Traps: Rational Misalignment Driven by Model Misspecification
di: Xu, Xingcheng, et al.
Pubblicazione: (2026)

Calibration Collapse Under Sycophancy Fine-Tuning: How Reward Hacking Breaks Uncertainty Quantification in LLMs
di: Sahoo, Subramanyam
Pubblicazione: (2026)

Do We Need Frontier Models to Verify Mathematical Proofs?
di: Naik, Aaditya, et al.
Pubblicazione: (2026)

Separating Constraint Compliance from Semantic Accuracy: A Novel Benchmark for Evaluating Instruction-Following Under Compression
di: Baxi, Rahul
Pubblicazione: (2025)

Adversarial Tokenization
di: Geh, Renato Lui, et al.
Pubblicazione: (2025)

On Evaluating The Performance of Watermarked Machine-Generated Texts Under Adversarial Attacks
di: Liu, Zesen, et al.
Pubblicazione: (2024)

Adversarial Attacks on AI-Generated Text Detection Models: A Token Probability-Based Approach Using Embeddings
di: Kadhim, Ahmed K., et al.
Pubblicazione: (2025)

Extrapolative Weight Averaging Reveals Correctness-Efficiency Frontiers in Code RL
di: Zheng, Kunhao, et al.
Pubblicazione: (2026)

MegaScience: Pushing the Frontiers of Post-Training Datasets for Science Reasoning
di: Fan, Run-Ze, et al.
Pubblicazione: (2025)

Generative Adversarial Reasoner: Enhancing LLM Reasoning with Adversarial Reinforcement Learning
di: Liu, Qihao, et al.
Pubblicazione: (2025)

Confidence Under the Hood: An Investigation into the Confidence-Probability Alignment in Large Language Models
di: Kumar, Abhishek, et al.
Pubblicazione: (2024)

An Adversarial Perspective on Machine Unlearning for AI Safety
di: Łucki, Jakub, et al.
Pubblicazione: (2024)

Resolving UnderEdit & OverEdit with Iterative & Neighbor-Assisted Model Editing
di: Baghel, Bhiman Kumar, et al.
Pubblicazione: (2025)

AceMath: Advancing Frontier Math Reasoning with Post-Training and Reward Modeling
di: Liu, Zihan, et al.
Pubblicazione: (2024)

The Metacognitive Probe: Five Behavioural Calibration Diagnostics for LLMs
di: Oliveira, Rafael C. T.
Pubblicazione: (2026)

From Instructions to Constraints: Language Model Alignment with Automatic Constraint Verification
di: Wang, Fei, et al.
Pubblicazione: (2024)

BookSQL: A Large Scale Text-to-SQL Dataset for Accounting Domain
di: Kumar, Rahul, et al.
Pubblicazione: (2024)

Recursive Introspection: Teaching Language Model Agents How to Self-Improve
di: Qu, Yuxiao, et al.
Pubblicazione: (2024)

The Growing Pains of Frontier Models: When Leaderboards Stop Separating and What to Measure Next
di: Amin, Adil
Pubblicazione: (2026)

Why Do Safety Guardrails Degrade Across Languages?
di: Zhang, Max, et al.
Pubblicazione: (2026)

Unfamiliar Finetuning Examples Control How Language Models Hallucinate
di: Kang, Katie, et al.
Pubblicazione: (2024)

The Reasoning Trap: An Information-Theoretic Bound on Closed-System Multi-Step LLM Reasoning
di: Shin, Kwan Soo
Pubblicazione: (2026)

The Flexibility Trap: Why Arbitrary Order Limits Reasoning Potential in Diffusion Language Models
di: Ni, Zanlin, et al.
Pubblicazione: (2026)

Impacts of Racial Bias in Historical Training Data for News AI
di: Bhargava, Rahul, et al.
Pubblicazione: (2025)