Datasets:
language:
- en
task_categories:
- text-generation
tags:
- reinforcement-learning
- verl
- opv
configs:
- config_name: train
data_files:
- split: train
path: train.parquet
- config_name: validation
data_files:
- split: validation
path: validation.parquet
- config_name: eurus_validation
data_files:
- split: eurus_validation
path: eurus_validation.parquet
OPV Code: original experiment data
Prepared for Learning to Steer, Steering to See. validation is the full local LiveCodeBench v5 export (880 problems), not the 100-row smoke subset. The original converter caps tests at 30 and their combined bytes; this is an execution-budget variant, not unrestricted official LCB evaluation.
| Split | Rows |
|---|---|
| train | 25,276 |
| validation | 880 |
| eurus_validation | 1,024 |
Provenance and terms
The data is derived from the following sources; their original terms and attribution obligations continue to apply. No new blanket license is asserted over the collection. Source-file and uploaded-file hashes are in manifest.json.
- https://huggingface.co/datasets/PRIME-RL/Eurus-2-RL-Data
- https://huggingface.co/datasets/livecodebench/code_generation_lite
Format and evaluation
These are byte-for-byte copies of the original Parquet files. All columns, Arrow schemas, nested verifier metadata, row order, prompts and answers are unchanged. Use the existing verl reward functions directly. Each file has a separate Hub configuration because original train/evaluation schemas differ.
Use the full evaluation split and four independently sampled responses per prompt; report mean@4, not pass@4. No training responses or model weights are included. Exact normalized-chat train/evaluation overlap: 0 prompts. Original splits are preserved, including any disclosed overlap, to match existing teacher provenance. This is not a semantic contamination audit.
from datasets import load_dataset
train = load_dataset("caiyuchen/OPV-Code", "train", split="train")
validation = load_dataset("caiyuchen/OPV-Code", "validation", split="validation")
The available artifacts differ from several manuscript descriptions. The accompanying repository records these differences instead of relabeling datasets or inventing results.