影响力指数

44.78/100

前 11.1%

全站排名 #7,158

发表论文12 篇

平均评分4.5

年均产出4.0 篇/年

Sumit Kumar Jha

Full Professor@University of Florida·美国·OpenReview

研究方向

mechanistic interpretation · agents · large language models · foundation models · diffusion · explainability · neural stochastic differential equations · neurosymbolic · logic · automata theory

5.0

Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion

ICLR 2026Poster

通讯

5.0

Hessian-Enhanced Token Attribution (HETA): Interpreting Autoregressive LLMs

ICLR 2026Poster

通讯

5.0

Verifiable Natural Language to Linear Temporal Logic Translation: A Benchmark Dataset and Evaluation Suite

ICLR 2026Rejected

4.7

Fact or Hallucination? An Entropy-Based Framework for Attention-Wise Usable Information in LLMs

ICLR 2026Withdrawn

通讯

3.2

Adaptive Text Transformations Defend Against Dataset Inference Attacks

ICLR 2026Rejected

三作

6.4

Fact or Hallucination? An Entropy-Based Framework for Attention-Wise Usable Information in LLMs

NeurIPS 2025Rejected

通讯

5.5

Enhancing Integrated Gradients Using Emphasis Factors and Attention for Effective Explainability of Large Language Models

合作者 (14)

Sumit Kumar Jha

Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion

Hessian-Enhanced Token Attribution (HETA): Interpreting Autoregressive LLMs

Verifiable Natural Language to Linear Temporal Logic Translation: A Benchmark Dataset and Evaluation Suite

Fact or Hallucination? An Entropy-Based Framework for Attention-Wise Usable Information in LLMs

Adaptive Text Transformations Defend Against Dataset Inference Attacks

Fact or Hallucination? An Entropy-Based Framework for Attention-Wise Usable Information in LLMs

Grammar-Forced Translation of Natural Language to Temporal Logic using LLMs

Metric-Driven Attributions for Vision Transformers

IMPROVING LOW-BIT POST TRAINING QUANTIZATION: A DATA-FREE APPROACH

Enhancing Integrated Gradients Using Emphasis Factors and Attention for Effective Explainability of Large Language Models