Shreyansh Modi

Reading List

A curated list of papers I have studied, each accompanied by my understanding of its key ideas and contributions.

Jailbreaks & Interpretability

• Toward Universal and Transferable Jailbreak Attacks on Vision-Language Models

  K. Cui, Y. Li, Y. Wu, X. Ma, S. Erfani, C. Leckie, H. Huang

  ICLR 2026

  My understanding

  To be added.

  Paper Code

• Prompt Injection as Role Confusion

  C. Ye, J. Cui, D. Hadfield-Menell

  ICML 2026

  My understanding

  To be added.

  Paper Project

• Mechanistic Interpretability of LLM Jailbreaks via Internal Attribution Graphs

  A. Wagle, I. I. Uddin, C. Zhang, L. Wang

  arXiv preprint, 2026

  My understanding

  To be added.

  Paper

• Adversarial Examples Are Not Bugs, They Are Features

  A. Ilyas, S. Santurkar, D. Tsipras, L. Engstrom, B. Tran, A. Madry

  NeurIPS 2019

  My understanding

  To be added.

  Paper

• Understanding Jailbreak Success: A Study of Latent Space Dynamics in Large Language Models

  S. Ball, F. Kreuter, N. Panickssery

  EACL 2026

  My understanding

  To be added.

  Paper

• Understanding and Rectifying Safety Perception Distortion in VLMs

  X. Zou, J. Kang, G. Kesidis, L. Lin

  NeurIPS 2025

  My understanding

  To be added.

  Paper

• Involuntary Jailbreak: On Self-Prompting Attacks

  Y. Guo, Y. Li, M. Kankanhalli

  TMLR 2026

  My understanding

  To be added.

  Paper

Alignment

• Safety Tax: Safety Alignment Makes Your Large Reasoning Models Less Reasonable

  T. Huang, S. Hu, F. Ilhan, S. F. Tekin, Z. Yahn, Y. Xu, L. Liu

  arXiv preprint, 2025

  My understanding

  To be added.

  Paper