Metacognition and the Source of Competence in Large Language Models
Roi Cohen et al. "Metacognition and the Source of Competence in Large Language Models." ICML 2026 Workshop. Extended version in preparation.
CV
Education
- Ph.D in Version Control Theory, GitHub University, 2018 (expected)
- M.S. in Jekyll, GitHub University, 2014
- B.S. in GitHub, GitHub University, 2012
Work experience
- Spring 2024: Academic Pages Collaborator
- GitHub University
- Duties includes: Updates and improvements to template
- Supervisor: The Users
- Fall 2015: Research Assistant
- GitHub University
- Duties included: Merging pull requests
- Supervisor: Professor Hub
- Summer 2015: Research Assistant
- GitHub University
- Duties included: Tagging issues
- Supervisor: Professor Git
Skills
- Skill 1
- Skill 2
- Sub-skill 2.1
- Sub-skill 2.2
- Sub-skill 2.3
- Skill 3
Publications
Knowledgeless Language Models: Decoupling Linguistic Competence and Factual Knowledge
Roi Cohen, Yvan Carre, Nick Lechtenborger, Hendrik Droste, Lucas Kerschke, Russa Biswas, Gerard de Melo, Jan Buys. "Knowledgeless Language Models: Decoupling Linguistic Competence and Factual Knowledge." Under review, NeurIPS 2026.
Uncertainty Beyond Geometry: Kernel-Based Functional Routing in LLMs
Roi Cohen, Omri Fahn, Gerard de Melo. "Uncertainty Beyond Geometry: Kernel-Based Functional Routing in LLMs." Under review, COLM 2026.
StyleAT: Defending Face Recognition Against Semantic Attacks
Ben Shapira, Roi Cohen, Shang-Tse Chen, Mahmood Sharif. "StyleAT: Defending Face Recognition Against Semantic Attacks." Under review, ICML 2026.
Unifying Latent Uncertainty Signals in Large Language Models for Improved Factual Precision
Roi Cohen, Omri Fahn, Gerard de Melo. "Unifying Latent Uncertainty Signals in Large Language Models for Improved Factual Precision." Under review, ICLR 2026.
InFact: Informativeness Alignment for Improved LLM Factuality
Roi Cohen, Russa Biswas, Gerard de Melo. "InFact: Informativeness Alignment for Improved LLM Factuality." Findings of EMNLP 2025.
I Don’t Know: Explicit Modeling of Uncertainty with an [IDK] Token
Roi Cohen, Konstantin Dobler, Eden Biran, Gerard de Melo. "I Don't Know: Explicit Modeling of Uncertainty with an [IDK] Token." NeurIPS 2024.
Evaluating the Ripple Effects of Knowledge Editing in Language Models
Roi Cohen, Eden Biran, Ori Yoran, Amir Globerson, Mor Geva. "Evaluating the Ripple Effects of Knowledge Editing in Language Models." TACL 2024.
LM vs LM: Detecting Factual Errors via Cross-Examination
Roi Cohen, May Hamri, Mor Geva, Amir Globerson. "LM vs LM: Detecting Factual Errors via Cross-Examination." EMNLP 2023.
Crawling the Internal Knowledge-Base of Language Models
Roi Cohen, Mor Geva, Jonathan Berant, Amir Globerson. "Crawling the Internal Knowledge-Base of Language Models." Findings of EACL 2023.
Talks
Conference Proceeding talk 3 on Relevant Topic in Your Field
Conference proceedings talk at Testing Institute of America 2014 Annual Conference, Los Angeles, CA, USA
Talk 2 on Relevant Topic in Your Field
Talk at London School of Testing, London, UK
Tutorial 1 on Relevant Topic in Your Field
Tutorial at UC-Berkeley Institute for Testing Science, Berkeley, CA, USA
Talk 1 on Relevant Topic in Your Field
Talk at UC San Francisco, Department of Testing, San Francisco, CA, USA
Teaching
Service and leadership
- Currently signed in to 43 different slack teams