Add pipeline tag, library name, and prominent GitHub link

This PR enhances the model card by adding key metadata and improving content organization:
- `pipeline_tag: text-generation`: This helps users discover the model when filtering by text generation models on the Hub, reflecting its function in confidence estimation for LLM outputs.
- `library_name: transformers`: This enables the automated "how to use" code snippet on the Hub, as the model's code clearly leverages the 🤗 Transformers library (e.g., `AutoModel.from_pretrained`, `PeftModel.from_pretrained`).
- A prominent link to the GitHub repository is added at the top of the README for easier access to the associated code and project details.

These additions will make the model more accessible and easier for users to integrate into their workflows.

README.md

@@ -1,19 +1,26 @@
 ---
-license: apache-2.0
-language:
-- en
 base_model:
 - Qwen/Qwen2.5-7B-Instruct
+language:
+- en
+license: apache-2.0
+pipeline_tag: text-generation
+library_name: transformers
 ---
-# Introduction
 
-This is the official repo of the paper [Annotation-Efficient Universal Honesty Alignment](https://arxiv.org/abs/2510.17509)
+# Annotation-Efficient Universal Honesty Alignment
+
+This is the official repository for the paper [Annotation-Efficient Universal Honesty Alignment](https://arxiv.org/abs/2510.17509).
+
+Code: [https://github.com/Trustworthy-Information-Access/Annotation-Efficient-Universal-Honesty-Alignment](https://github.com/Trustworthy-Information-Access/Annotation-Efficient-Universal-Honesty-Alignment)
+
+## Introduction
 
 This repository provides modules that extend **Qwen2.5-7B-Instruct** with the ability to generate accurate confidence scores *before* response generation, indicating how likely the model is to answer a given question correctly across tasks. We offer two types of modules—**LoRA + Linear Head** and **Linear Head**—along with model parameters under three training settings:
 
-1. **Elicitation (greedy):** Trained on all questions (over 560k) using self-consistency-based confidence annotations.
-2. **Calibration-Only (right):** Trained on questions with explicit correctness annotations.
-3. **EliCal (hybrid):** Initialized from the Elicitation model and further trained on correctness-labeled data.
+1.  **Elicitation (greedy):** Trained on all questions (over 560k) using self-consistency-based confidence annotations.
+2.  **Calibration-Only (right):** Trained on questions with explicit correctness annotations.
+3.  **EliCal (hybrid):** Initialized from the Elicitation model and further trained on correctness-labeled data.
 
 For both **Calibration-Only** and **EliCal** settings, we provide models trained with different amounts of annotated data (1k, 2k, 3k, 5k, 8k, 10k, 20k, 30k, 50k, 80k, 200k, 560k+). Since **LoRA + Linear Head** is the main configuration used in our paper, the following description is based on this setup.