ReVeR-Traces-Rubric

Supervised fine-tuning data for the rubric generator of ReVeR. Each of the 30,310 examples maps an instruction to an evaluation rubric, together with the teacher's reasoning. Every rubric item is labelled [Hard Rule] or [Principle].

The companion dataset ReVeR-Traces-Judge holds the judge trajectories for the same rubrics. The two datasets are aligned by id: every rubric here is used, word for word, by exactly two trajectories there, one for each response of a preference pair.

How the data was built

1. For each preference pair (an instruction, a preferred response and a rejected response), a teacher model (DeepSeek-V4-Flash with reasoning) wrote a rubric from the instruction alone.
2. The same teacher judged each response separately against that rubric: 0 or 1 for a Hard Rule, an integer from 0 to 100 for a Principle.
3. A pair is kept only if the judge scores rank the preferred response strictly higher. The score of a response is the number of satisfied Hard Rules plus the mean Principle score divided by 101, so Principles only break ties between responses that satisfy the same number of Hard Rules.

Fields

Field	Content
id	Id of the preference pair. It joins the example to its two trajectories in ReVeR-Traces-Judge
source	Origin of the pair: helpsteer3, ultrafeedback, skywork, magpie or ifeval (synthetic instruction-following prompts)
messages	Two messages. user: the rubric-generation prompt with the instruction. assistant: the teacher's reasoning in <think>…</think>, then one <rubric> block

Usage

from datasets import load_dataset

ds = load_dataset("OpenRubrics/ReVeR-Traces-Rubric", split="train")

example = ds[0]
prompt = example["messages"][0]["content"]
target = example["messages"][1]["content"]

# The reasoning may mention the tag, so take the last <rubric> block.
start = target.rfind("<rubric>") + len("<rubric>")
rubric = target[start:target.rfind("</rubric>")].strip()

Citation

If you use this dataset, please cite our paper Rubrics from the Task, Verdicts from Evidence: Pointwise Rubric Rewards with Tool-Augmented Judges:

@misc{liu2026rubricstaskverdictsevidence,
      title={Rubrics from the Task, Verdicts from Evidence: Pointwise Rubric Rewards with Tool-Augmented Judges}, 
      author={Tianci Liu and Haoyu Wang},
      year={2026},
      eprint={2610.18766},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2610.18766}, 
}