Add dataset card, paper link, and sample usage

README.md

@@ -36,4 +36,43 @@ configs:
   data_files:
   - split: train
     path: data/train-*
+task_categories:
+- text-generation
+language:
+- en
 ---
+
+# Small Reward Models via Backward Inference (FLIP)
+
+This dataset is used for **FLIP** (FLipped Inference for Prompt reconstruction), a reference-free and rubric-free reward modeling approach introduced in the paper [Small Reward Models via Backward Inference](https://huggingface.co/papers/2602.13551).
+
+- **Paper:** [Small Reward Models via Backward Inference](https://huggingface.co/papers/2602.13551)
+- **GitHub:** [yikee/FLIP](https://github.com/yikee/FLIP)
+
+## Dataset Description
+The dataset contains approximately 12k English prompts derived from the WildChat dataset. It is designed to support reinforcement learning training (such as via GRPO), where the reward is calculated by inferring the instruction from a model's response and comparing it to the original ground-truth instruction.
+
+## Sample Usage
+
+To compute the reward using the F1 score as described in the paper (requires the `metrics.py` file from the official repository):
+
+```python
+from metrics import f1_score
+
+# prediction = inferred instruction from the LLM
+# ground_truth = original instruction
+result = f1_score(prediction, ground_truth)["f1"]
+```
+
+The `metrics.py` utility also provides `normalize_answer(s)` for normalizing text before comparison (lowercasing, removing punctuation/articles, and fixing whitespace).
+
+## Citation
+
+```latex
+@article{wang2026small,
+  title={Small Reward Models via Backward Inference},
+  author={Wang, Yike and Brahman, Faeze and Feng, Shangbin and Xiao, Teng and Hajishirzi, Hannaneh and Tsvetkov, Yulia},
+  journal={arXiv preprint arXiv:2602.13551},
+  year={2026}
+}
+```