Writerslogic at PAN 2026: Process over Content for Robust Detection under Domain Shift
arXiv cs.CL · 3 Oct 2026, 02:44 AEST
Writerslogic presented their systems for three PAN at CLEF 2026 shared tasks, emphasizing process over content for robust detection under domain shift. Their analytical framework highlights the importance of feature robustness and explains the performance outcomes across various tasks.
Key facts
- We describe the Writerslogic systems for three PAN at CLEF 2026 shared tasks (Reasoning Trajectory Detection, Voight-Kampff Generative AI Detection, and Multi-Author Writing Style Analysis), unified by a shared analytical framework: feature robustness under distribution shift is governed by support overlap between training and test distributions, not by training-set effect size.
- On Reasoning Trajectory Detection, where training was entirely mathematics and 84 percent of test was unseen domains, the framework guided system design to 1st place in source detection (0.85 macro F1 via Opus-Sonnet agreement) and 3rd place in safety classification (0.66 macro F1 via query-refusal decomposition).
- For Voight-Kampff, we built a calibrated ensemble of DeBERTa-v2 (ONNX), multi-seed LightGBM with 44 domain-portable stylometric features, and SVM on n-gram TF-IDF, combined via learned stacking with isotonic calibration; the best configuration achieved 0.891 on the PAN 2026 test set with balanced sub-metrics (0.853 to 0.902 across all evaluation dimensions).
- Across all three tasks, features measuring generation process properties are designed to outperform features measuring generated content properties under domain shift.
Read aloud
Writerslogic presented their systems for three PAN at CLEF twenty twenty six shared tasks, emphasizing process over content for robust detection under domain shift. Their analytical framework highlights the importance of feature robustness and explains the performance outcomes across various tasks. We describe the Writerslogic systems for three PAN at CLEF twenty twenty six shared tasks Reasoning Trajectory Detection, Voight-Kampff Generative A I Detection, and Multi-Author Writing Style Analysis, unified by a shared analytical framework: feature robustness under distribution shift is governed by support overlap between training and test distributions, not by training-set effect size. On Reasoning Trajectory Detection, where training was entirely mathematics and eighty four percent of test was unseen domains, the framework guided system design to 1st place in source detection zero point eight five macro F1 via Opus-Sonnet agreement and 3rd place in safety classification zero point six six macro F1 via query-refusal decomposition. For Voight-Kampff, we built a calibrated ensemble of DeBERTa-v2 ONNX, multi-seed LightGBM with forty four domain-portable stylometric features, and SVM on n-gram TF-IDF, combined via learned stacking with isotonic calibration; the best configuration achieved zero point eight nine one on the PAN twenty twenty six test set with balanced sub-metrics zero point eight five three to zero point nine zero two across all evaluation dimensions. Across all three tasks, features measuring generation process properties are designed to outperform features measuring generated content properties under domain shift.
Sources in this cluster
- Writerslogic at PAN 2026: Process over Content for Robust Detection under Domain Shift
arXiv cs.CL · 3 Oct 2026, 02:44 AEST