Obtaining faithful interpretations from compositional neural networks

Sanjay Subramanian, Ben Bogin, Nitish Gupta, Tomer Wolfson, Sameer Singh, Jonathan Berant, Matt Gardner

نتاج البحث: فصل من :كتاب / تقرير / مؤتمرمنشور من مؤتمرمراجعة النظراء

ملخص

Neural module networks (NMNs) are a popular approach for modeling compositionality: they achieve high accuracy when applied to problems in language and vision, while reflecting the compositional structure of the problem in the network architecture. However, prior work implicitly assumed that the structure of the network modules, describing the abstract reasoning process, provides a faithful explanation of the model's reasoning; that is, that all modules perform their intended behaviour. In this work, we propose and conduct a systematic evaluation of the intermediate outputs of NMNs on NLVR2 and DROP, two datasets which require composing multiple reasoning steps. We find that the intermediate outputs differ from the expected output, illustrating that the network structure does not provide a faithful explanation of model behaviour. To remedy that, we train the model with auxiliary supervision and propose particular choices for module architecture that yield much better faithfulness, at a minimal cost to accuracy.

اللغة الأصليةالإنجليزيّة
عنوان منشور المضيفACL 2020 - 58th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference
ناشرAssociation for Computational Linguistics (ACL)
الصفحات5594-5608
عدد الصفحات15
رقم المعيار الدولي للكتب (الإلكتروني)9781952148255
حالة النشرنُشِر - 2020
الحدث58th Annual Meeting of the Association for Computational Linguistics, ACL 2020 - Virtual, Online, الولايات المتّحدة
المدة: ٥ يوليو ٢٠٢٠١٠ يوليو ٢٠٢٠

سلسلة المنشورات

الاسمProceedings of the Annual Meeting of the Association for Computational Linguistics

!!Conference

!!Conference58th Annual Meeting of the Association for Computational Linguistics, ACL 2020
الدولة/الإقليمالولايات المتّحدة
المدينةVirtual, Online
المدة٥/٠٧/٢٠١٠/٠٧/٢٠

All Science Journal Classification (ASJC) codes

  • !!Computer Science Applications
  • !!Linguistics and Language
  • !!Language and Linguistics

بصمة

أدرس بدقة موضوعات البحث “Obtaining faithful interpretations from compositional neural networks'. فهما يشكلان معًا بصمة فريدة.

قم بذكر هذا