Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models
cs.AI, cs.CL, cs.SE
Submitted: 2026-09-24
Updated: 2026-09-24
Code: https://github.com/ebarkhordar/llm-collusion
Terminology
Sources
- Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators
- Inspection and Control of Self-Generated-Text Recognition Ability in Llama3-8b-Instruct
- StyleRemix: Interpretable Authorship Obfuscation via Distillation and Perturbation of Style Elements
- Program Synthesis with Large Language Models
- Know Thyself? On the Incapability and Implications of AI Self-Recognition
- When can we trust untrusted monitoring? A safety case sketch across collusion strategies
- Looking Inward: Language Models Can Learn About Themselves by Introspection
- AI Control: Improving Safety Despite Intentional Subversion
- I Know Which LLM Wrote Your Code Last Summer: LLM generated Code Stylometry for Authorship Attribution
- Code Fingerprints: Disentangled Attribution of LLM-Generated Code
- DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation
- Can AI Recognize Its Own Reflection? Self-Detection Performance of LLMs in Computing Education
- Me, Myself, and AI: The Situational Awareness Dataset (SAD) for LLMs
- AI Content Self-Detection for Transformer-based Large Language Models
- Hidden in Plain Text: Emergence & Mitigation of Steganographic Collusion in LLMs
- Evaluating Large Language Models Trained on Code
- Your Large Language Models Are Leaving Fingerprints
- Self-Recognition in Language Models
- Don't Judge Code by Its Cover: Exploring Biases in LLM Judges for Code Evaluation
- Self-Preference Bias in LLM-as-a-Judge
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection