← Flash PapersSpurious Rewards: Rethinking Training Signals in RLVR

Share this Flash Paper

Spurious Rewards: Rethinking Training Signals in RLVR

https://flashpapers.ai/p/bnb9crsych
Post to XPost to BlueskyShare on LinkedInSubmit to Hacker NewsSubmit to RedditEmail a link
Make your own
Rulin Shao, Shuyue Stella Li +12 more
cs.AI2025arXiv:2506.10947Code
Details ▸
MATH-500 Accuracy:70.8%· MATH-500Qwen2.5-Math-7B MATH-500 Accuracy Gain (Random Rewards):+21.4 percentage points· MATH-500Qwen2.5-Math-7B MATH-500 Accuracy Gain (Ground Truth):+29.1 percentage points· MATH-500Code Reasoning Frequency:95.6%· MATH-500