Add link to paper and project page

README.md

@@ -1,25 +1,28 @@
 ---
-pretty_name: Agent Retrieval Bench
-license: other
 language:
 - en
+license: other
+size_categories:
+- n<1K
 task_categories:
 - text-retrieval
+pretty_name: Agent Retrieval Bench
 tags:
 - code
 - software-engineering
 - code-retrieval
 - coding-agents
 - agentic-retrieval
-size_categories:
-- n<1K
 ---
+
 # Agent Retrieval Bench
 
 [![Dataset](https://img.shields.io/badge/Hugging%20Face-dataset-yellow)](https://huggingface.co/datasets/eyuansu71/agent_retrieval_bench)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 [![Citation](https://img.shields.io/badge/citation-CFF-lightgrey)](CITATION.cff)
 
+[Paper](https://huggingface.co/papers/2607.24882) | [Project page](https://agent-retrieval-bench.github.io/) | [Code](https://github.com/eyuansu62/agent-retrieval-bench)
+
 Agent Retrieval Bench (ARB) evaluates the context-acquisition layer of coding
 agents: given a real workflow signal and a repository at a frozen base commit,
 can a retriever find the files an agent needs to read next, or correctly decide
@@ -123,6 +126,7 @@ Positive retrieval results over all 345 samples:
 | RepoMap | 0.6333 | 0.2158 | **0.3788** |
 | nomic-embed-code | 0.5244 | 0.1986 | 0.2781 |
 | jina-code-embeddings-0.5b | 0.4823 | 0.1914 | 0.2783 |
+| Nomic | 0.5244 | 0.1986 | 0.2781 |
 | Lexical | 0.4940 | 0.1574 | 0.2650 |
 | BM25 | 0.4452 | 0.1520 | 0.2051 |
 
@@ -348,4 +352,4 @@ code, benchmark metadata, and documentation are MIT licensed; see
 }
 ```
 
-A machine-readable citation record is available in [CITATION.cff](CITATION.cff).
+A machine-readable citation record is available in [CITATION.cff](CITATION.cff).