影响力指数

16.08/100

超过 48.6%

全站排名 #33,075

发表论文3 篇

平均评分4.1

年均产出1.0 篇/年

Philip Quirke

Researcher@Martian·美国·OpenReview

研究方向

AI Safety · AI Mechanistic Interpretability

Understanding Addition and Subtraction in Transformers

ICLR 2026Rejected

Planning in a recurrent neural network that plays Sokoban

ICLR 2025Rejected

合作者 (8)

Aaron David Tucker

Adrià Garriga-Alonso

Mohammad Taufeeque