The First International Competition on Computational Models of Argumentation (ICCMA'15) will be conducted in the first half of 2015. The results will be presented at the workshop "Theory and Applications of Formal Argument" (TAFA'15) which will be co-located with the 24th International Joint Conference on Artificial Intelligence (IJCAI'15).

ICCMA'15 will focus on reasoning tasks in abstract argumentation frameworks, cf. [Dung:1995]. Submitted solvers will be tested on several artificially generated argumentation frameworks and some frameworks formalizing real-world problems.

For the first instance of the competition we will evaluate solvers based on their performance in solving the following computational tasks:

1. Given an abstract argumentation framework, determine some extension
2. Given an abstract argumentation framework, determine all extensions
3. Given an abstract argumentation framework and some argument, decide whether the given argument is credulously inferred
4. Given an abstract argumentation framework and some argument, decide whether the given argument is skeptically inferred

The above computational tasks are to be solved with respect to the following standard semantics:

1. Complete Semantics
2. Preferred Semantics
3. Grounded Semantics
4. Stable Semantics

Developers of solvers may decide to only provide support for a subset of the above computational tasks and/or semantics. For each category (task+semantics) we will provide a separate ranking of the submitted solvers.

The competition will be conducted using the open source benchmark suite probo.

News:

• 2017/08/31: A detailed report on the competition of 2015 has been published in the Artificial Intelligence Journal.
• 2016/04/14: A summary report on the competition has been published in the AI Magazine.
• 2015/10/20: A volume collecting all system descriptions of solvers submitted to ICCMA'15 has been published at arXiv: http://arxiv.org/abs/1510.05373.
• 2015/09/30: The statistics of the results of ICCMA'15 have been updated. Thanks to Fabio Rossi for spotting some errors in the representation of conarg.
• 2015/09/27: The benchmark generators of ICCMA'15 are now available in the source code repository of probo.
• 2015/08/11: The competition is over and the results are in. Check out the result page to get detailed statistics on the evaluation of the solvers.
• 2015/07/07: System descriptions of submitted solvers now available.
• 2015/04/14: List of submitted solvers now available
• 2015/03/17: Added requirements for system description submission./li>
• 2015/03/05: Updated version of iccma15notes_v3.pdf (corrects a file format typo).
• 2015/02/03: We are quite happy to announce that we received 16 declarations of intent so far!

• 2014/12/22: As an early christmas present, we uploaded a second set of benchmark examples. Our benchmark generator has been tweaked a bit and the new set contains more interesting examples than the first set (e.g. more and different extensions).

  We also provide the ground truth for each argumentation graph in the archive under the folder "solutions". For each semantics (CO, PR, GR, ST) and each argument graph, there is one file in "solutions" containing the set of all extensions wrt. the given semantics. We also implemented a solver that reads the solutions from these files in order to compare the output directly in probo.

  You can download the archive with the new benchmarks and their solutions here (the first number of the filename is an indicator for the hardness of the graph: the larger the harder). We also uploaded the solutions to the first benchmark set at (only solutions to the "rdm1000" test cases are missing).

  Happy holidays!

• 2014/11/21: We uploaded an archive containing some example graphs in TGF and APX format to the competition website (zip).

  We invite all potential participants to try out these examples with their solver. The graphs are drawn from different probabilistic models with different sizes and there are also some graphs taken from real-world debates. For now, we will not disclose the exact parameters on how the random graphs are generated in order to avoid an optimisation of solvers towards specific graph models. However, in terms of structure these example graphs are representative for the graphs that will actually be used in the competition.

  As for the size of the graphs, we included in the archive graphs ranging from 20 arguments to more than 1000 arguments. As ICCMA'15 will be the first of its kind, we would be very happy if you can tell us a bit how your solvers perform on the different sizes of graphs. We currently estimate that we will set a time-out of 10-30 minutes for each computational problem on each graph (e.g. computing all preferred extensions on one graph). It would be very good if you can give us some feedback about how large the graphs can be while still being solved in the given time-out (please send your feedback to sc@argumentationcompetition.org).

  As for the file formats, we currently only support TGF and APX (see iccma15notes_v3.pdf for a description) and removed support of the CNF format as there are ambiguities when converting CNF to the other formats. If you wish to have another format supported, please do not hesitate to express your wish on our mailinglist argumentationcompetition@inria.fr. We can easily add support for any other format.

• 2014/11/21: New version of the general interface description of ICCMA'15 (pdf)
• 2014/11/18: New version of tutorial on how to implement the solver interface for ICCMA'15 (pdf)
• 2014/11/13: Updated tutorial on how to implement the solver interface for ICCMA'15 (pdf), see also the general interface description (pdf)