初始化项目，由ModelHub XC社区提供模型

Model: sebsigma/SemanticCite-Refiner-Qwen3-1B
Source: Original Platform

Modelfile

@@ -0,0 +1,29 @@
+FROM hf.co/sebsigma/SemanticCite-Refiner-Qwen3-1B
+
+PARAMETER temperature 0.6
+PARAMETER top_p 0.95
+PARAMETER top_k 40
+PARAMETER num_ctx 1024
+PARAMETER repeat_penalty 1.1
+PARAMETER stop "<|im_end|>"
+PARAMETER stop "### Input:"
+PARAMETER stop "### Response:"
+
+TEMPLATE """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
+
+### Instruction:
+{{ .System }}
+
+### Input:
+{{ .Prompt }}
+
+### Response:
+"""
+
+SYSTEM """Process the input citation by:
+1. Removing any reference markers, author names, publication identifiers or citations (e.g., [1], (Smith, 2020), et al.)
+2. Converting author-centered statements to fact-centered statements (often using passive voice)
+3. Maintaining all numerical values, specific metrics, and factual details
+4. Ensuring the claim stands alone as a verifiable statement
+
+Return only the processed claim without any explanation or additional text."""