ORACLE

Workshop on Open Reasoning Across Cultures & Languages

Reasoning lies at the core of human intelligence: it allows us to connect entities, events, and quantities into coherent arguments. While the underlying logical capacity is universal, the ways in which people reason are shaped by the languages they speak and the cultural practices they adopt—whether in how numbers are conceptualised, how space is oriented, or how categories are formed.

Generative AI, particularly LLMs, has made great progress in complex reasoning, but far less attention has been paid to extending these abilities across languages. Progress remains uneven: while English dominates, other languages reveal systematic gaps in the representation of typological variation, script change, and cultural context. When extended beyond English, performance degrades under typological variation, code-switching, and data scarcity. Furthermore, evaluation protocols often lack cross-lingual fidelity and reproducibility, especially in low-resource settings where linguistic and cultural diversity most acutely impedes progress.

ORACLE represents a forum for discussing new ideas and directions to advance Open Reasoning Across Cultures & Languages through transparent methods, reusable resources, and reproducible evaluation. By foregrounding transparency, openness, and reproducibility, ORACLE aims to catalyse research outcomes capable of interpreting and solving complex multilingual problems, paving the way for systems that support reliable, culturally aware reasoning at scale.

Call for Papers

We welcome contributions of previously unpublished papers, which could be either long (8 pages) or short (4 pages). All submissions will be peer-reviewed by multiple reviewers. The authors' identities must be concealed to enable a double-blind peer review.

ORACLE welcomes both archival and non-archival submissions. Only archival submissions will be included in the proceedings.

Our focus is on methods, data, and evaluations that advance Open Reasoning Across Languages & Cultures.

We invite work that promotes novel contributions with an emphasis on:

Open data, evaluation suites, and tools.
Rigorous, reproducible solutions to broader multilingual capabilities.

We are particularly interested in (but not limited to) works related to the following topics:

Evaluation methods in multilingual spaces (including cultural grounding and robustness).
Domain-grounded multilingual applications (science, education, finance, healthcare).
Agentic workflows and autonomous reasoning across linguistic and cultural boundaries.
Resources for inferring culturally mediated meanings.
Building trust in multilingual GenAI through faithful, well-calibrated, and interpretable outputs.
Ensuring compliance, safety, and ethical alignment in multicultural AI deployments.

Submissions can be made via OpenReview.

Please, use the official ACL template.