影响力指数

70.35/100

前 2.6%

全站排名 #1,668

发表论文14 篇

平均评分5.6

年均产出4.7 篇/年

Naomi Saphra

Assistant Professor@Boston University, Boston University·美国·OpenReview

研究方向

animal communication · mode connectivity · loss landscapes · compositionality · learning dynamics · training dynamics · interpretability · language models · natural language processing

5.5

Hidden Breakthroughs in Language Model Training

ICLR 2026Poster

三作

5.0

Do Natural Language Descriptions of Model Activations Convey Privileged Information?

ICLR 2026Rejected

4.0

Can Interpretation Predict Behavior on Unseen Data?

ICLR 2026Withdrawn

通讯

3.3

Random Scaling of Emergent Capabilities

ICLR 2026Rejected

通讯

6.8

Recite, Reconstruct, Recollect: Memorization in LMs as a Multifaceted Phenomenon

ICLR 2025Poster

通讯

6.5

PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs

合作者 (20)

Naomi Saphra

Hidden Breakthroughs in Language Model Training

Do Natural Language Descriptions of Model Activations Convey Privileged Information?

Can Interpretation Predict Behavior on Unseen Data?

Random Scaling of Emergent Capabilities

Recite, Reconstruct, Recollect: Memorization in LMs as a Multifaceted Phenomenon

PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs

A Taxonomy of Transcendence

Sometimes I am a Tree: Data Drives Unstable Hierarchical Generalization

Loss in the Crowd: Hidden Breakthroughs in Language Model Training

Structural Probing with Feature Interaction