The Answer Lies Within: Self-Derived Rewards Enable Explainable Relation Extraction
Xinyu Guo, Zhengliang Shi, Minglai Yang, Mahdi Rahimi, Mihai Surdeanu
EMNLP 2026, main track
About
I work on natural language processing, that is, building computer systems that process and extract meaning from natural language texts. I focus mostly on interpretable models, i.e., approaches where the computer can explain in human understandable terms why it made a decision. I am particularly interested in neuro-symbolic approaches that combine neural systems, such as large language models, with interpretable symbolic methods, such as deterministic rules.
Please see our lab's web site for more information on our research, current and former students, etc.
Latest news
- August 2026 — Two papers accepted in the main track at EMNLP and one in TMLR!
- August 2026 — I am on sabbatical during the 2026/2027 academic year. I won't take any new students during this time.
- December 2025 - I received a Fulbright award for my coming sabbatical!
Research interests
- Natural language processing
- Explainable AI
- Neuro-symbolic AI
- Machine reasoning
- Information extraction
- Biomedical NLP
Selected publications
AlignSAE: Concept-Aligned Sparse Autoencoders
Minglai Yang, Xinyu Guo, Zhengliang Shi, Jinhe Bi, Steven Bethard, Mihai Surdeanu, Liangming Pan
Transactions on Machine Learning Research (TMLR), 2026
Beyond Sequence Order: Syntax-Informed Positional Embeddings for Transformers
Haris Riaz, Hyungji Kim, Mihai Surdeanu
EMNLP 2026, main track
For all our papers please see our lab's publication page or my Google Scholar.
Students
Please see our lab's web site for information on our current and former students.
Teaching
- [Spring 2026] CSC 483/583: Text Retrieval and Web Search
- [Fall 2025] CSC 396: Introduction to Deep Learning with Applications to Natural Language Processing