LongPuzzleBench
Copyright 2026 LongPuzzleBench contributors

The evaluation harness contains code adapted from MobileWorld
(https://github.com/Tongyi-MAI/MobileWorld), revision
e222907a5709ed20053847a2dca5f63c283177cc, under the Apache License 2.0.
LongPuzzleBench adds the browser-game runtime, puzzle task catalog, progressive
execution protocol, benchmark scoring, result aggregation, and bundled puzzle
environments. The upstream Android environments, private evaluation services,
and deployment infrastructure are not included.

The bundled puzzle suite was integrated from the hongbin branch of the
Bolt_Unscrew project (https://github.com/adf1178/Bolt_Unscrew/tree/hongbin),
based on revision 282ad0b, and extended for benchmark use.
See games/puzzle_suite/NOTICE.md and THIRD_PARTY_NOTICES.md.