Urmăriți
Thomas Dooms
Thomas Dooms
Adresă de e-mail confirmată pe uantwerpen.be
Titlu
Citat de
Citat de
Anul
The Trifecta: Three simple techniques for training deeper Forward-Forward networks
T Dooms, IJ Tsang, J Oramas
TMLR, 2023
22023
Bilinear MLPs enable weight-based mechanistic interpretability
MT Pearce*, T Dooms*, A Rigg, J Oramas, L Sharkey
arXiv preprint arXiv:2410.08417, 2024
12024
Weight-based Decomposition: A Case for Bilinear MLPs
MT Pearce, T Dooms, A Rigg
ICML 2024 Workshop on Mechanistic Interpretability, 2024
12024
Compositionality Unlocks Deep Interpretable Models
T Dooms*, W Gauderis*, G Wiggins, JAO Mogrovejo
Connecting Low-Rank Representations in AI: At the 39th Annual AAAI …, 2024
2024
Tokenized SAEs: Disentangling SAE Reconstructions
T Dooms*, D Wilhelm*
ICML 2024 Workshop on Mechanistic Interpretability, 2024
2024
Sistemul nu poate realiza operația în acest moment. Încercați din nou mai târziu.
Articole 1–5