Product
Help
Pricing
About
Jobs
Log in
Get started
Author
Add to collection
Stefan Heimersheim
Observed name:
Heimersheim, Stefan
Institutions:
Harvard University,
University of Toronto
Observed institutions:
Harvard University,
University of Toronto
OpenAlex ID:
A5114412487
Towards Automated Circuit Discovery for Mechanistic Interpretability
2023
·
Arthur Conmy
, Augustine Mavor-Parker
, et al.
·
neural information processing systems
·
258
PDF
How to use and interpret activation patching
2024
·
Heimersheim, Stefan
, Nanda, Neel
·
arXiv (Cornell University)
·
238
PDF
Evolution of SAE Features Across Layers in LLMs
2024
·
Balcells, Daniel
, Lerner, Benjamin
, et al.
·
arXiv (Cornell University)
·
14
PDF
Concept Influence: Leveraging Interpretability to Improve Performance and Efficiency in Training Data Attribution
2026
·
Kowal, Matthew
, Paulo, Goncalo
, et al.
·
arXiv (Cornell University)
·
9
PDF
The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes
2026
·
Taufeeque, Mohammad
, Heimersheim, Stefan
, et al.
·
arXiv (Cornell University)
·
8
PDF
You can remove GPT2's LayerNorm by fine-tuning
2024
·
Heimersheim, Stefan
·
arXiv (Cornell University)
·
8
PDF
The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology
2026
·
Szablewski, Andrzej
, Konar-Steenberg, Gabriel
, et al.
·
arXiv (Cornell University)
·
6
PDF
Characterizing stable regions in the residual stream of LLMs
2024
·
Janiak, Jett
, Karwowski, Jacek
, et al.
·
arXiv (Cornell University)
·
6
PDF
Evaluating Synthetic Activations composed of SAE Latents in GPT-2
2024
·
Giglemiani, Giorgi
, Petrova, Nora
, et al.
·
arXiv (Cornell University)
·
4
PDF
Evidence for feature-specific error correction in LLMs
2026
·
da Silva, Francisco Ferreira
, Heimersheim, Stefan
·
arXiv (Cornell University)
·
1
PDF
Investigating Sensitive Directions in GPT-2: An Improved Baseline and Comparative Analysis of SAEs
2024
·
Lee, Daniel J.
, Heimersheim, Stefan
·
arXiv (Cornell University)
·
1
PDF
Individual Parameters in Weight-Sparse Transformers Appear Interpretable
2026
·
Marin-Llobet, Arnau
, Heimersheim, Stefan
·
arXiv (Cornell University)
PDF
Show more works
View as search
Works count:
12
Citation count:
553
H-index:
6
I10-index:
3
Year
Topic
Type
