The 2nd LLMs4OL Challenge consists of 4 tasks:
- Task A - Text2Onto: Extract ontological terminologies and types from a raw text.
- Task B - Term Typing: Discover the generalized type for a lexical term.
- Task C - Taxonomy Discovery: Discover the taxonomic hierarchy between type pairs.
- Task D - Non-Taxonomic Relation Extraction: Identify non-taxonomic, semantic relations between types.
Two main evaluation phases for the challenge are:
- Seen-Eval Testing Phase – Each selected ontology is split into training and testing subsets. Participants train their systems on the training portion, while the reserved portion is used for evaluation in this phase.
- Blind-Eval Testing Phase – This phase assesses generalizability by introducing new, unseen ontologies for testing, ensuring that models can adapt to novel data beyond their training set.