Aryaman Arora · आर्यमन अरोरा
I am a Member of Technical Staff on the Interpretability team at Anthropic. I want to understand how language models work.
Before this, I defended my Ph.D. at Stanford University, advised by Dan Jurafsky and Christopher Potts, and funded by the NSF Graduate Research Fellowship Program. Before that, I completed my B.S. in Computer Science and Linguistics at Georgetown University, where I worked with Nathan Schneider on computational linguistics.
I have interned at ETH Zürich (with Ryan Cotterell), Apple, Redwood Research, and Transluce.
Contact
Google Scholar · GitHub · Twitter · Email
Greatest Hits [» more papers]
Matryoshka attribution: Learning to attribute language model
outputs to representations and weights
PreFT: Prefill-only finetuning for inference efficiency
ADAG: Automatically describing attribution graphs
Language model circuits are sparse in the neuron basis
Mechanistic evaluation of Transformers and state space
models
AxBench: Steering LLMs? Even simple baselines outperform sparse
autoencoders
ReFT: Representation finetuning for language models
CausalGym: Benchmarking causal interpretability methods on
linguistic tasks
Travels