Dharshan Kumaran

Researcher, Google

Joined

December 2017

Names

Dharshan Kumaran

Emails

****@google.com (Confirmed)

Personal Links

Career & Education History

Researcher

Google (deepmind.com)

2014 – 2034

Advisors, Relations & Conflicts

Postdoc Advisor

James Lloyd McClelland

2009 – 2014

Expertise

Metacognition

2024 – 2034

Confidence

2024 – 2034

Neuroscience

2003 – 2034

Publications

How LLMs Detect and Correct Their Own Errors: The Role of Internal Confidence Signals
Dharshan Kumaran, Viorica Patraucean, Simon Osindero, Petar Veličković, Nathaniel D. Daw
- arxiv
- Readers: Everyone
Causal Evidence that Language Models use Confidence to Drive Behavior
Dharshan Kumaran, Nathaniel Daw, Simon Osindero, Petar Velickovic, Viorica Patraucean
- CoRR 2026
- Readers: Everyone
How do LLMs Compute Verbal Confidence
Dharshan Kumaran, Arthur Conmy, Federico Barbero, Simon Osindero, Viorica Patraucean, Petar Velickovic
- CoRR 2026
- Readers: Everyone
How Overconfidence in Initial Choices and Underconfidence Under Criticism Modulate Change of Mind in Large Language Models
Dharshan Kumaran, Stephen M. Fleming, Larisa Markeeva, Joseph Heyward, Mrinal Mathur, Andrea Banino, Razvan Pascanu, Simon Osindero, Benedetto De Martino, Petar Veličković, Viorica Patraucean
- Arxiv
- Readers: Everyone
How Overconfidence in Initial Choices and Underconfidence Under Criticism Modulate Change of Mind in Large Language Models
Dharshan Kumaran, Stephen M. Fleming, Larisa Markeeva, Joe Heyward, Andrea Banino, Mrinal Mathur, Razvan Pascanu, Simon Osindero, Benedetto De Martino, Petar Velickovic, Viorica Patraucean
- CoRR 2025
- Readers: Everyone
Language models, like humans, show content effects on reasoning tasks
Andrew Kyle Lampinen, Ishita Dasgupta, Stephanie C.Y. Chan, Hannah Sheahan, Antonia Creswell, Dharshan Kumaran, James Lloyd McClelland, Felix Hill
- PNAS Nexus
- Readers: Everyone
Transformers need glasses! Information over-squashing in language tasks
Federico Barbero, Andrea Banino, Steven Kapturowski, Dharshan Kumaran, João Guilherme Madeira Araújo, Alex Vitvitskyi, Razvan Pascanu, Petar Veličković
- TF2M 2024 Poster
- Readers: Everyone
Transformers need glasses! Information over-squashing in language tasks
Federico Barbero, Andrea Banino, Steven Kapturowski, Dharshan Kumaran, João Guilherme Madeira Araújo, Alex Vitvitskyi, Razvan Pascanu, Petar Veličković
- NeurIPS 2024 poster
- Readers: Everyone
Transformers need glasses! Information over-squashing in language tasks
Federico Barbero, Andrea Banino, Steven Kapturowski, Dharshan Kumaran, João G. M. Araújo, Alex Vitvitskyi, Razvan Pascanu, Petar Velickovic
- CoRR 2024
- Readers: Everyone
Transformers need glasses! Information over-squashing in language tasks
Federico Barbero, Andrea Banino, Steven Kapturowski, Dharshan Kumaran, João Guilherme Madeira Araújo, Oleksandr Vitvitskyi, Razvan Pascanu, Petar Velickovic
- NeurIPS 2024
- Readers: Everyone

View all 27 publications

Co-Authors

View all 126 co-authors