Can We Trust LLM Detectors?
This addresses the problem of unreliable AI text detection for users relying on LLM-generated content, but it is incremental as it builds on existing paradigms.
The paper systematically evaluates existing LLM text detectors, finding they are brittle under distribution shifts and stylistic perturbations, and proposes a supervised contrastive learning framework to learn discriminative style embeddings, though it highlights fundamental challenges in building domain-agnostic detectors.
The rapid adoption of LLMs has increased the need for reliable AI text detection, yet existing detectors often fail outside controlled benchmarks. We systematically evaluate 2 dominant paradigms (training-free and supervised) and show that both are brittle under distribution shift, unseen generators, and simple stylistic perturbations. To address these limitations, we propose a supervised contrastive learning (SCL) framework that learns discriminative style embeddings. Experiments show that while supervised detectors excel in-domain, they degrade sharply out-of-domain, and training-free methods remain highly sensitive to proxy choice. Overall, our results expose fundamental challenges in building domain-agnostic detectors. Our code is available at: https://github.com/HARSHITJAIS14/DetectAI