Record link: 
http://hdl.handle.net/20.500.12708/226143
-
Title: 
Using Action-Policy Testing in RL to Reduce the Number of Bugs
en
Citation: 
Eniser, H. F., Lin, S., Müller, N., Isychev, A., Wüstholz, V., Valera, I., Hoffmann, J., & Christakis, M. (2025). Using Action-Policy Testing in RL to Reduce the Number of Bugs. In M. Likhachev, H. Rudová, & E. Scala (Eds.), Eighteenth International Symposium on Combinatorial Search (pp. 181–185). AAAI Press. http://hdl.handle.net/20.500.12708/226143
-
Publication Type: 
Inproceedings - Full-Paper Contribution
en
Language: 
English
-
Authors: 
Eniser, Hasan Ferit 
Lin, Songtuan 
Müller, Nicola 
Isychev, Anastasia 
Wüstholz, Valentin 
Valera, Isabel 
Hoffmann, Jörg 
Christakis, Maria 
-
Organisational Unit: 
E194-01 - Forschungsbereich Software Engineering 
E056-26 - Fachbereich Automated Reasoning 
-
Published in: 
Eighteenth International Symposium on Combinatorial Search 
-
ISBN: 
978-1-57735-901-2
-
Date (published): 
20-Jul-2025
-
Event name: 
The 18th International Symposium on Combinatorial Search (SoCS 2025) 
en
Event date: 
12-Aug-2025 - 15-Aug-2025
-
Event place: 
Glasgow, United Kingdom of Great Britain and Northern Ireland (the)
-
Number of Pages: 
5
-
Publisher: 
AAAI Press, Washington DC
-
Peer reviewed: 
Yes
-
Keywords: 
action-policy testing; reinforcement learning; bug reduction
en
Abstract: 
Reinforcement learning is becoming ever more prominent in solving combinatorial search problems, in particular ones where states are images. Prior work has devised action-policy testing methodology, that identifies so-called bug states where policy performance is sub-optimal. Here we show how to leverage this methodology during the RL process, using action-policy testing to find bugs and injecting those as alternate start states for the training runs. Running experiments across six 2D games, we find that our testing-guided training often achieves similar expected reward while reducing the number of bugs.
en
Research Areas: 
Information Systems Engineering: 100%
-
Science Branch: 
1020 - Informatik: 100%
-
Appears in Collections:
Conference Paper

Show full item record

Page view(s)

37
checked on Feb 9, 2026

Google ScholarTM

Check