ABD: Default Exception Abduction in Finite First Order Worlds
Fuente:
arXiv
Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914526941675520 |
|---|---|
| author | Batzoglou, Serafim |
| author_facet | Batzoglou, Serafim |
| contents | We introduce ABD, a benchmark for default-exception abduction over finite first-order worlds. Given a background theory with an abnormality predicate and a set of relational structures, a model must output a first-order formula that defines exceptions, restoring satisfiability while keeping exceptions sparse. We formalize three observation regimes (closed-world, existential completion, universal completion) with exact SMT verification. Evaluating ten frontier LLMs on 600 instances, the best models achieve high validity but parsimony gaps remain, and holdout evaluation reveals distinct generalization failure modes across regimes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_18843 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ABD: Default Exception Abduction in Finite First Order Worlds Batzoglou, Serafim Artificial Intelligence Symbolic Computation We introduce ABD, a benchmark for default-exception abduction over finite first-order worlds. Given a background theory with an abnormality predicate and a set of relational structures, a model must output a first-order formula that defines exceptions, restoring satisfiability while keeping exceptions sparse. We formalize three observation regimes (closed-world, existential completion, universal completion) with exact SMT verification. Evaluating ten frontier LLMs on 600 instances, the best models achieve high validity but parsimony gaps remain, and holdout evaluation reveals distinct generalization failure modes across regimes. |
| title | ABD: Default Exception Abduction in Finite First Order Worlds |
| topic | Artificial Intelligence Symbolic Computation |
| url | https://arxiv.org/abs/2602.18843 |