From 4bdb8c7d6b9d57a881fd134c5d10934df35e384b Mon Sep 17 00:00:00 2001 From: Oliver Date: Tue, 25 Feb 2025 22:39:06 +0000 Subject: [PATCH] Add note on code execution to CodeIODataset --- reasoning_gym/code/codeio.py | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/reasoning_gym/code/codeio.py b/reasoning_gym/code/codeio.py index 105c480e..c0fa3e54 100644 --- a/reasoning_gym/code/codeio.py +++ b/reasoning_gym/code/codeio.py @@ -63,6 +63,12 @@ class CodeIOConfig: class CodeIODataset(ProceduralDataset): + """ + Exercise some caution when using this dataset, as it involves executing arbitrary code snippets. + These code snippets are transformed by an LLM from raw code files which have been curated from high-quality sources. + However, there is still a risk that the LLM could have introduced code with bad effects. + """ + _jsonl_data: Optional[list] = None def __init__(self, config: CodeIOConfig):