Scholar
Andrew Kyle Lampinen
Google Scholar ID: _N44XxAAAAAJ
Research Scientist, DeepMind
Citations & Impact
Citations
5,066
H-index
26
i10-index
33
Publications
20
Co-authors
39
list available
Publications
Resume (English only)
Academic Achievements
- Demonstrated how in-context learning in language models emerges from simple data distributional properties.
- Showed that language models, like humans, exhibit content effects in logical reasoning.
- Revealed how model representations are biased by irrelevant properties, posing challenges for interpretability and neuroscience.
- Explored aligning vision model representations with human semantic knowledge to improve ML robustness and cognitive modeling.
- Investigated how explanations and interventional structures in internet data enable language models to learn causality from passive data.
- Worked on scaling grounded language agents across virtual environments to enhance compositional generalization.
