open-thought/reasoning-gym
1
0
Fork 0
mirror of https://github.com/open-thought/reasoning-gym.git synced 2026-04-22 16:49:06 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Constrain reward

Browse source
This commit is contained in:
Oliver 2025-02-17 19:20:45 +00:00
parent 1d0cad46f2
commit eb708e88b3
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
reasoning_gym/code/bf.py
View file

@ -127,10 +127,10 @@ int main() {{
if entry["answer"] in answer.splitlines():
# We can be quite confident that the correct answer was given
# It was likely just given alongside an explanation
return 0.9 * len(answer) / len(entry["answer"])
return max(0.9 * len(answer) / len(entry["answer"]), 0.1)
if entry["answer"] in answer:
# Since answers are English words, some risk of the response coincidentally containing the answer
return 0.5 * len(answer) / len(entry["answer"])
return max(0.5 * len(answer) / len(entry["answer"]), 0.1)
return 0.01
else:
return 1.0 # Yay