Add dataset card and metadata for DocQA-RL-1.6K

README.md

@@ -0,0 +1,31 @@
+---
+task_categories:
+- question-answering
+license: apache-2.0
+tags:
+- long-context
+- reasoning
+- reinforcement-learning
+- question-answering
+- docqa
+---
+
+# DocQA-RL-1.6K: A Reinforcement Learning Dataset for Long-Context Question Answering
+
+This dataset, `DocQA-RL-1.6K`, comprises 1.6K document question answering (DocQA) problems designed for training and evaluating large reasoning models (LRMs) on long-context reasoning tasks. The dataset is divided into three reasoning domains:
+
+1. **Mathematical Reasoning:** 600 problems from the DocMath dataset, requiring numerical reasoning across long and specialized documents.
+2. **Logical Reasoning:** 600 multi-choice questions synthesized using DeepSeek-R1, requiring logic analysis of real-world documents.
+3. **Multi-Hop Reasoning:** 400 examples (200 from MultiHopRAG and 200 from Musique), emphasizing cross-document reasoning.
+
+This dataset was used in the training and evaluation of the QwenLong-L1 model, as described in the paper: [QwenLong-L1: Towards Long-Context Large Reasoning Models with Reinforcement Learning](https://huggingface.co/papers/2505.17667).
+
+**Links:**
+
+* **Github:** [https://github.com/Tongyi-Zhiwen/QwenLong-L1](https://github.com/Tongyi-Zhiwen/QwenLong-L1)
+
+**Related Datasets (for evaluation):**
+
+* docmath (https://huggingface.co/datasets/Tongyi-Zhiwen/docmath)
+* frames (https://huggingface.co/datasets/Tongyi-Zhiwen/frames)
+* longbench (https://huggingface.co/datasets/Tongyi-Zhiwen/longbench)