Research map / Large language models / Instruction and fine-tuning
$f$-Divergence Self-Play for Tabular Anomaly Detection via Large Language Models
Hoang Tran Vuong, Linh Ngo Van, Dang Nguyen, Thin Nguyen, Phuoc Nguyen, Mehrtash Harandi, Trung Le
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace one-shot SFT (AnoLLM) with f-divergence self-play | AnoLLM SmolLM-135M · Six-dataset benchmark | AUC-ROC (%): 76.5% -> 84.6% | Improved |
| replace one-shot SFT (AnoLLM) with f-divergence self-play | AnoLLM SmolLM-360M · Six-dataset benchmark | AUC-ROC (%): 77.8% -> 85.4% | Improved |
| replace one-shot SFT (AnoLLM) with f-divergence self-play | AnoLLM SmolLM-1.7B · Six-dataset benchmark | AUC-ROC (%): 77.7% -> 83.5% | Improved |
| replace one-shot SFT (AnoLLM) with f-divergence self-play | AnoLLM SmolLM-1.7B · Six-dataset benchmark | F1 (%): 40.1% -> 48.0% | Improved |
| replace one-shot SFT (AnoLLM) with f-divergence self-play | AnoLLM SmolLM-1.7B · Fakejob | AUC-PR (%): 31.2% -> 30.0% | Hurt |
| replace one-shot SFT (AnoLLM) with f-divergence self-play | AnoLLM SmolLM-360M · Vehicle | F1 (%): 17.4% -> 19.6% | Hurt |
| add self-play iterations (up to iter-3) after SFT (ablation) | DiSPaT SmolLM-135M · Fakejob, Vehicle, 20 Newsgroups, Ecoli | AUC-ROC | Mixed |
| use KL conjugate e^v-1 instead of identity (SPIN) conjugate (ablation) | DiSPaT SmolLM-135M · 4 datasets avg | AUC-ROC (%): 83.7% -> 85.7% | Improved |
| use KL conjugate e^v-1 instead of identity (SPIN) conjugate (ablation) | DiSPaT SmolLM-135M · 4 datasets avg | F1 (%): 44.9% -> 52.7% | Improved |
| use KL conjugate e^v-1 instead of identity (SPIN) conjugate (ablation) | DiSPaT SmolLM-135M · 4 datasets avg | AUC-PR (%): 44.6% -> 50.5% | Improved |
| use reverse KL instead of identity conjugate (ablation) | DiSPaT SmolLM-135M · 4 datasets avg | AUC-ROC (%): 83.7% -> 84.4% | Improved |
| use squared Hellinger instead of identity conjugate (ablation) | DiSPaT SmolLM-135M · 4 datasets avg | AUC-ROC (%): 83.7% -> 84.8% | Improved |
Closest papers by meaning
- Beyond Individual Input for Deep Anomaly Detection on Tabular Data (ICML 2024 · similarity 0.88)
- Disentangling Tabular Data Towards Better One-Class Anomaly Detection (AAAI 2025 · similarity 0.87)
- Beyond Normal References: Discriminative Few-Shot Anomaly Detection (NeurIPS 2026 · similarity 0.85)
- TabDPT: Scaling Tabular Foundation Models on Real Data (NeurIPS 2025 · similarity 0.85)
- Tab-Shapley: Identifying Top-k Tabular Data Quality Insights (AAAI 2025 · similarity 0.85)
- TabDiff: a Mixed-type Diffusion Model for Tabular Data Generation (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
