Warm congratulations to Dr. Martin Tutek, postdoctoral researcher under the supervision of Dr. Yonatan Belinkov, whose paper “Measuring Faithfulness of Chains of Thought by Unlearning Reasoning Steps” received the Outstanding Paper Award at the prestigious EMNLP 2025 conference.
The paper, co-authored with Dr. Yonatan Belinkov and collaborators from the University of Utah (USA), addresses the reliability and faithfulness of AI systems based on large language models.
The researchers propose a novel method for evaluating the trustworthiness of multi-step reasoning explanations by analyzing the effect of each reasoning step on the model’s internal computations. This approach advances our ability to assess the faithfulness of AI systems and may inform future improvements in their design and reliability.
The award was presented last week at the EMNLP 2025 conference held in Suzhou, China.
Read the paper: https://aclanthology.org/2025.emnlp-main.504.pdf
[Back to the news index]