EUVIMEDEUVIMED entdecken
Crossref

Learning from Reasoning Failures via Synthetic Data Generation

Gabriela Ben Melech Stan, Estelle Aflalo, Avinash Madasu, Vasudev Lal, Phillip Howard

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (30) · 25608-25616

DOI: 10.1609/aaai.v40i30.39757 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Gabriela Ben Melech Stan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Training models on synthetic data has emerged as an increasingly important strategy for improving the performance of generative AI. This approach is particularly helpful for large multimodal models (LMMs) due to the relative scarcity of high-quality paired image-text data compared to language-only data. While a variety of methods have been proposed for generating large multimodal datasets, they do not tailor the synthetic data to address specific deficiencies in the reasoning abilities of LMMs which will be trained with the generated dataset. In contrast, humans often learn in a more efficient manner by seeking out examples related to the types of reasoning where they have failed previously. Inspired by this observation, we propose a new approach for synthetic data generation which is grounded in the analysis of an existing LMM's reasoning failures. Our methodology leverages frontier models to automatically analyze errors produced by a weaker LMM and propose new examples which can be used to correct the reasoning failure via additional training, which are then further filtered to ensure high quality. We generate a large multimodal instruction tuning dataset containing over 553k examples using our approach and conduct extensive experiments demonstrating its utility for improving the performance of LMMs on multiple downstream tasks. Our results show that models trained on our synthetic data can even exceed the performance of LMMs trained on an equivalent amount of additional real data, demonstrating the high value of generating synthetic data targeted to specific reasoning failure modes in LMMs.

Zitieren

APA

Stan, G. B. M., Aflalo, E., Madasu, A., Lal, V., & Howard, P. (2026). Learning from Reasoning Failures via Synthetic Data Generation. Proceedings of the AAAI Conference on Artificial Intelligence, 40(30), 25608-25616. https://doi.org/10.1609/aaai.v40i30.39757

Vancouver

Stan GBM, Aflalo E, Madasu A, Lal V, Howard P. Learning from Reasoning Failures via Synthetic Data Generation. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(30):25608-25616. doi:10.1609/aaai.v40i30.39757

BibTeX
@article{Stan2026,
  author = {Stan, Gabriela Ben Melech and Aflalo, Estelle and Madasu, Avinash and Lal, Vasudev and Howard, Phillip},
  title = {Learning from Reasoning Failures via Synthetic Data Generation},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {30},
  pages = {25608--25616},
  doi = {10.1609/aaai.v40i30.39757}
}
Kostenlos starten →