LuxVeri at GenAI Detection Task 3: Cross-Domain Detection of AI-Generated Text Using Inverse Perplexity-Weighted Ensemble of Fine-Tuned Transformer Models
This addresses the problem of detecting AI-generated text across domains for content moderation applications, but it is incremental as it builds on existing transformer methods with modest performance gains.
The paper tackled cross-domain detection of AI-generated text by proposing an ensemble of fine-tuned transformer models with inverse perplexity weighting, achieving TPR scores of 0.826 (ranking 10th/23) for non-adversarial detection and 0.801 (ranking 8th/22) for adversarial detection.
This paper presents our approach for Task 3 of the GenAI content detection workshop at COLING-2025, focusing on Cross-Domain Machine-Generated Text (MGT) Detection. We propose an ensemble of fine-tuned transformer models, enhanced by inverse perplexity weighting, to improve classification accuracy across diverse text domains. For Subtask A (Non-Adversarial MGT Detection), we combined a fine-tuned RoBERTa-base model with an OpenAI detector-integrated RoBERTa-base model, achieving an aggregate TPR score of 0.826, ranking 10th out of 23 detectors. In Subtask B (Adversarial MGT Detection), our fine-tuned RoBERTa-base model achieved a TPR score of 0.801, securing 8th out of 22 detectors. Our results demonstrate the effectiveness of inverse perplexity-based weighting for enhancing generalization and performance in both non-adversarial and adversarial MGT detection, highlighting the potential for transformer models in cross-domain AI-generated content detection.