Defeasible benchmark that can be used to classify defeasible reasoning tools based on their semantics (ambiguity propagating vs ambiguity blocking), expressiveness (existential rules, negation, cyclic rules, priority relation) and performance.
It is recommended to run the benchmark using command line, it is also possible to run it directly. See BenchLaucher.java.
- Ambiguity: Tests if the tool has an ambiguity blocking or an ambiguity propagating behavior.
- Team Defeat (Direct Reinstatement): Tests if the tool allows for team defeat or not.
- Floating Conclusions: Tests if the tool accepts floating conclusions.
- Consistent Answers: Tests if the tool considers conflicts that appear after applying strict rules.
- Existential Test: Test if the tool understands existential rules.
- Chain FES: Test performance when faced with FES rules.
- Chain FUS: Test performance with FUS rules.
- Chain GBTS: Test performance with GBTS rules.
- Cyclic GRD: Test performance when the GRD is cyclic.
- Cyclic Conflict: Test performance when faced with cyclic conflicts.
- Circular Reasoning: Tests if the tool considers circular reasoning.
- Rule Application Block: Test performance when defeaters or negated label on rules are used to block their application.
- Priority: Test performance w.r.t. a priority relation.
- Queries: Test performance when faced with non ground queries.
- Chain Theory: Test performance with long chains of rules.
- Tree Theory: Test performance w.r.t. a large number of linked arguments.
- Directed Acyclic Graph Theory: Test performance when faced with many arguments for the same atom.
Implement the interface AbstractTool. Example: DeLPTool.
This project is licensed under the MIT License.