Publications

Also see the selected publications on the home page. Author lists are abbreviated with “et al.” past the first few names on large collaborations; Tutek, M. is bolded throughout. Entries marked [TODO] are ones I couldn’t confidently verify a link or acceptance status for — flagging rather than guessing.

2026

• CRISP: Persistent Concept Unlearning via Sparse Autoencoders
  Ashuach, T., Arad, D., Mueller, A., Tutek, M., Belinkov, Y.
  ACL 2026 paper
• Old Habits Die Hard: How Conversational History Geometrically Traps LLMs
  Simhi, A., Barez, F., Tutek, M., Belinkov, Y., Cohen, S. B.
  ICML 2026 paper
• Reasoning Models Know What's Important, and Encode It in Their Activations
  Nikankin, Y., Tutek, M., Ashuach, T., Rosenfeld, J. S., Belinkov, Y.
  COLM 2026 paper
• Sequence Repetition Enhances Token Embeddings and Improves Sequence Labeling with Decoder-only Language Models
  Kukić, M. L., Čuljak, M., Dukić, D., Tutek, M., Šnajder, J.
  Findings of EACL 2026 paper
• PragWorld: A Benchmark Evaluating LLMs' Local World Model under Minimal Linguistic Alterations and Conversational Dynamics
  Vashistha, S., Bibhuti, A., Naik, A., Tutek, M., Aditya, S.
  AAAI 2026 (oral) paper
• ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs
  Simhi, A., Herzig, J., Tutek, M., Itzhak, I., Szpektor, I., Belinkov, Y.
  ICLR 2026 paper & data arXiv
• Evaluating Pluralism in LLMs through Latent Perspectives
  Majer, L., Šnajder, J., Tutek, M.
  Pluralistic Alignment Workshop @ ICML 2026 paper
• Context Parametrization with Compositional Adapters
  Jukić, J., Tutek, M., Šnajder, J.
  arXiv preprint [TODO: confirm venue — under review at ICLR 2026 as of Oct 2025 CV] paper

2025

• Measuring Faithfulness of Natural Language Explanations by Unlearning Reasoning Steps
  Tutek, M., Hashemi Chaleshtori, F., Marasović, A., Belinkov, Y.
  EMNLP 2025 🏆 Outstanding Paper Award paper
• MIB: A Mechanistic Interpretability Benchmark
  Mueller, A.*, Geiger, A.*, Wiegreffe, S.*, et al., Tutek, M., et al., Belinkov, Y.
  ICML 2025 paper website
• REVS: Unlearning Sensitive Information in Language Models via Rank Editing in the Vocabulary Space
  Ashuach, T., Tutek, M., Belinkov, Y.
  Findings of ACL 2025 paper code & website
• Characterizing Linguistic Shifts in Croatian News via Diachronic Word Embeddings
  Dukić, D., Barić, A., Čuljak, M., Jukić, J., Tutek, M.
  Slavic NLP Workshop @ ACL 2025 paper
• Findings of the BlackboxNLP 2025 Shared Task: Localizing Circuits and Causal Variables in Language Models
  Arad, D., Belinkov, Y., Chen, H., Kim, N., Mohebbi, H., Mueller, A., Sarti, G., Tutek, M.
  BlackboxNLP Workshop @ EMNLP 2025 (shared task findings) paper
• Predicting Success of Model Editing Through Intrinsic Features
  Soker, Y., Tutek, M., Belinkov, Y.
  Interplay Workshop @ COLM 2025 paper

2024

• Code Prompting Elicits Conditional Reasoning Abilities in Text+Code LLMs
  Puerto, H., Tutek, M., Aditya, S., Zhu, X., Gurevych, I.
  EMNLP 2024 paper
• Out-of-Distribution Detection by Leveraging Between-Layer Transformation Smoothness
  Jelenić, F., Jukić, J., Tutek, M., Puljiz, M., Šnajder, J.
  ICLR 2024 [TODO: swap for the OpenReview link] paper
• CATfOOD: Counterfactual Augmented Training for Improving Out-of-Domain Performance and Calibration
  Sachdeva, R., Tutek, M., Gurevych, I.
  EACL 2024 paper

2023

• Easy to Decide, Hard to Agree: Reducing Disagreements Between Saliency Methods
  Jukić, J.*, Tutek, M.*, Šnajder, J. (*equal contribution)
  Findings of ACL 2023 paper

2022

• Toward Practical Usage of the Attention Mechanism as a Tool for Interpretability
  Tutek, M., Šnajder, J.
  IEEE Access, 2022 [TODO: add link — no accessible DOI found]
• NLPOP: A Dataset for Popularity Prediction of Promoted NLP Research on Twitter
  Obadić, L., Tutek, M., Šnajder, J.
  WASSA Workshop @ ACL 2022 paper

2020

• Staying True to Your Word: (How) Can Attention Become Explanation?
  Tutek, M., Šnajder, J.
  RepL4NLP Workshop @ ACL 2020 paper

2018

• Iterative Recursive Attention Model for Interpretable Sequence Classification
  Tutek, M., Šnajder, J.
  BlackboxNLP Workshop @ EMNLP 2018 paper

2017

• Predicting News Values from Headline Text and Emotions
  di Buono, M. P., Šnajder, J., Dalbelo Bašić, B., Glavaš, G., Tutek, M., Milić-Frayling, N.
  NLP Meets Journalism Workshop @ EMNLP 2017 paper
• TakeLab at SemEval-2017 Task 5: Linear Aggregation of Word Embeddings for Fine-Grained Sentiment Analysis of Financial News
  Rotim, L., Tutek, M., Šnajder, J.
  SemEval Workshop @ ACL 2017 paper
• Two Layers of Annotation for Representing Event Mentions in News Stories
  di Buono, M. P., Tutek, M., Šnajder, J., Glavaš, G., Dalbelo Bašić, B., Milić-Frayling, N.
  Linguistic Annotation Workshop @ EACL 2017 paper

2016

• Detecting and Ranking Conceptual Links between Texts Using a Knowledge Base
  Tutek, M., Glavaš, G., Šnajder, J., Milić-Frayling, N., Dalbelo Bašić, B.
  CIKM 2016 [TODO: add link — paywalled on the ACM Digital Library]
• TakeLab at SemEval-2016 Task 6: Stance Classification in Tweets Using a Genetic Algorithm Based Ensemble
  Tutek, M., Sekulić, I., Gombar, P., Paljak, I., Čulinović, F., Boltužić, F., Karan, M., Alagić, D., Šnajder, J.
  SemEval Workshop @ NAACL 2016 paper