The results of the 2021 MaxSAT Evaluation were just announced at the 2021 SAT conference (by me :)). My solver Loandra performed very well, taking first place in the weighted 300s track and fourth in all other tracks. The source code of that version of Loandra is already available at the evaluation website and in its repository.
I would like to thank my co-organizers of the Evaluation and the organizers of the SAT conference. I am looking forward to taking a closer look at the results.
The MaxSAT Evaluation is an annual event meant to evaluate the state of the art in MaxSAT solving technology. The evaluation has two tracks: the complete track, where the goal is to find the provably optimal solution as fast as possible and the incomplete track where the goal is to find a solution which is as good as possible within a limited time.
This year we submitted a solver called Loandra to the incomplete track of the competition. Loandra implements core boosted linear search, an idea described in this paper that I worked on during my visit to Melbourne last fall. The results of the Evaluations were presented yesterday at the SAT conference, I’ve only seen a summary but I am happy to say that Loandra performed very well, achieving the highest score on the unweighted 300s track and the second to highest on the weighted 300s track. Overall a very nice result, I am very pleased.
I’d like to thank the organizers of the MaxSAT evaluation as well as my co-authors, Emir Demirovic and Peter Stuckey! I am currently on a holiday, but once I get back I will put up Loandra on github as well. It will also eventually be made available here.
EDIT: Loandra is now available at: https://github.com/jezberg/loandra and on the software pages.