Joseph An
I'm a BS/MS student in the Allen School at the University of Washington working on LLM reasoning and quantization research. Currently, I'm building agent infrastructure for clinical trials at Astraea. In Winter 2027, I'll be joining Quadrillion to work on autonomous research agents.
papers
-
Encode Once, Decode Never: Reusing Audio LM Internals for Efficient Temporal Localization ↗
COLM 2026
other things
- latent-cot ↗ llm reasoning with superposed token embeddings
- neuron-finder ↗ finding monosemantic neurons in qwen
- tpu-inference ↗ assorted patches to vllm's tpu backend