Add model card for AnomaMind

README.md

@@ -0,0 +1,47 @@
+---
+license: mit
+library_name: transformers
+pipeline_tag: other
+tags:
+- time-series
+- anomaly-detection
+- agent
+---
+
+# AnomaMind: Agentic Time Series Anomaly Detection with Tool-Augmented Reasoning
+
+AnomaMind is an agentic time series anomaly detection (TSAD) framework that reformulates anomaly detection as an evidence-driven sequential decision-making process. Unlike static methods, AnomaMind utilizes a structured coarse-to-fine workflow to localize suspicious intervals, construct diagnostic evidence through tool interaction, and refine decisions through self-reflection.
+
+- **Paper:** [AnomaMind: Agentic Time Series Anomaly Detection with Tool-Augmented Reasoning](https://huggingface.co/papers/2602.13807)
+- **Repository:** [Xiaoyu-Tao/AnomaMind-TS](https://github.com/Xiaoyu-Tao/AnomaMind-TS)
+
+## Model Description
+
+AnomaMind operates through a three-stage workflow:
+1. **Coarse Evidence Acquisition**: Localizes suspicious intervals using visual-aware models.
+2. **Adaptive Evidence Construction**: Uses a Detection Toolkit (statistical, value-based, and structural operators) to gather measurable evidence for verification.
+3. **Reasoning-based Detection & Refinement**: Employs a hybrid inference mechanism where general-purpose LLMs handle tool orchestration and self-reflection, while a detection-specific policy (this model) is optimized via reinforcement learning for precise decision-making.
+
+## Usage
+
+As per the official repository, you can serve the model using [vLLM](https://github.com/vllm-project/vllm) to provide an OpenAI-compatible API for the agentic workflow:
+
+```bash
+# Replace the path with your checkpoint path
+vllm serve ./model/PATH/TO/HUGGINGFACE --port 8000 --max-model-len 11000 --gpu-memory-utilization 0.95 --enable-auto-tool-choice --tool-call-parser hermes --served-model-name detector
+```
+
+Once served, the model can be used with the `infer.py` script provided in the GitHub repository to perform agentic anomaly detection.
+
+## Citation
+
+If you find AnomaMind useful in your research, please cite:
+
+```bibtex
+@article{tao2026anomamind,
+  title={AnomaMind: Agentic Time Series Anomaly Detection with Tool-Augmented Reasoning},
+  author={Tao, Xiaoyu and Wu, Yuchong and Cheng, Mingyue and Guo, Ze and Gao, Tian},
+  journal={arXiv preprint arXiv:2602.13807},
+  year={2026}
+}
+```