影响力指数
论文质量、代表作、近期表现、广度与样本量置信度综合计算
45.22/100
前 10.8%
全站排名 #6,976
发表论文8 篇
平均评分
年均产出4.0 篇/年
Constantin Venhoff
研究方向
Mechanistic Interpretability · Vision Language Models · AI Safety · Natural Language Processing
12
Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learning
ICLR 2026Poster
二作14
Dyslexify: A Mechanistic Defense Against Typographic Attacks in CLIP
ICLR 2026Poster
二作9
Towards Understanding Multimodal Fine-Tuning: A Case Study into Spatial Features
ICLR 2026Rejected
通讯6
Base Models Know How to Reason, Thinking Models Learn When
ICLR 2026Withdrawn
一作8
Predicting Weak-to-Strong Generalization from Latent Representations
ICLR 2026Rejected
通讯