Update README.md

README.md

@@ -1,3 +1,17 @@
----
-license: cc-by-nc-sa-4.0
----
+---
+license: cc-by-nc-sa-4.0
+---
+Hi all the soccer fans and all those who are interested in our project of **"Multi-Agent System for Comprehensive Soccer Understanding"**, here is the official source of our proposed benchmark **SoccerBench**. This benchmark is the largest and most comprehensive soccer-specific benchmark, featuring around 10K standardized multimodal (text, image, video) multi-choice QA pairs across 13 distinct understanding tasks, curated through automated pipelines and manual verification.
+
+More details of this benchmark and entire project could be refered from the following links:
+[Paper](https://arxiv.org/pdf/2505.03735) | [Code](https://github.com/jyrao/SoccerAgent) | [Arxiv](https://arxiv.org/abs/2505.03735)
+
+<figure>
+    <img src="https://jyrao.github.io/SoccerAgent/static/images/benchmark.png" alt="Proposed Benchmark Overview" width="80%">
+    <figcaption><strong>Proposed Benchmark Overview.</strong> SoccerBench QA Generation Pipeline. We construct multi-choice QA samples based on SoccerWiki and other existing datasets. Some representative examples for each task are presented for reference.</figcaption>
+</figure>
+
+<figure>
+    <img src="https://jyrao.github.io/SoccerAgent/static/images/benchmark_table.png" alt="Data Statistics of SoccerBench" width="80%">
+    <figcaption><strong>Data Statistics of SoccerBench.</strong> For each, we present its name, QA type, source materials, and curation strategies. Here, SN and SR-1988 represent the SoccerNet and Soccer-Replay-1988, respectively, while LLM denotes DeepSeek-v3.</figcaption>
+</figure>