Metraipolis › archive › dc_f87aa1d94fae

Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study — arXiv:2610.38326

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.38326
License (as archived)
CC0-1.0
Archived
2026-10-31T06:08:52.618Z
SHA-256 of the archived text
6b6059bf5c4c2681f86cd439a28509ad09155387c11978d52be67bbbbcc8b61a
Stable URL of this page
https://metraipolis.com/documento/dc_f87aa1d94fae

How to cite

Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study — arXiv:2610.38326 — arXiv (en), https://arxiv.org/abs/2610.38326 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f87aa1d94fae (Metraipolis archive dc_f87aa1d94fae, sha256 6b6059bf5c4c2681f86cd439a28509ad09155387c11978d52be67bbbbcc8b61a)
BibTeX
@misc{dcf87aa1d94fae,
  title = {Can LLMs Faithfully Solve Ontology Engineering Tasks? A Mechanistic Interpretability Study — arXiv:2610.38326},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.38326},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f87aa1d94fae; sha256 6b6059bf5c4c2681f86cd439a28509ad09155387c11978d52be67bbbbcc8b61a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.38326. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Ontological knowledge underpins large language model (LLM) performance on knowledge-intensive tasks. However, existing evaluations of LLMs' ontological understanding largely rely on behavioral correctness, leaving the underlying mechanisms unexplained. Existing mechanistic interpretability work has largely focused on instance-level factual knowledge, leaving ontological knowledge under-explored. To bridge this gap, we introduce MIOE, a causal-abstraction benchmark that evaluates interchange interventions for two ontology engineering tasks: entity typing and subsumption prediction. MIOE tests whether learned internal intervention targets can change an ontology-level target answer while preserving class-specific auxiliary information. Across Qwen3-8B and Qwen3-4B, we evaluate SAE-, DBM-, and DAS-based methods, including multi-task and multi-layer variants. Distributed subspace methods give the strongest results and generalize across prompt settings, but the induced changes remain only partially separated from auxiliary class-specific information. Multi-task objectives and contiguous layer-window interventions improve this balance, while class-level analyses reveal substantial variation across ontology classes. These results suggest that ontological predictions are mechanistically accessible in LLMs, although current methods do not yet isolate a cleanly disentangled causal mechanism for ontological knowledge.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.38326 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-31T06:07:08.470Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).