mirror of
https://github.com/AI-secure/UDora.git
synced 2026-02-12 21:03:21 +00:00
update readme
This commit is contained in:
@@ -1,6 +1,6 @@
|
||||
<div align="center">
|
||||
|
||||
<img src="https://github.com/AI-secure/UDora/assets/icon.png" width="200" alt="UDora Logo">
|
||||
<img src="https://github.com/AI-secure/UDora/blob/main/assets/icon.png" width="200" alt="UDora Logo">
|
||||
|
||||
# 🎯 [ICML 2025] UDora: A Unified Red Teaming Framework against LLM Agents
|
||||
|
||||
@@ -17,7 +17,7 @@
|
||||
|
||||
## 🔥 Pipeline Overview
|
||||
|
||||

|
||||

|
||||
|
||||
*UDora's dynamic attack pipeline: (1) Initial reasoning collection, (2) Weighted interval scheduling for optimal noise placement, (3) Adversarial string optimization, (4) Iterative refinement based on agent feedback.*
|
||||
|
||||
|
||||
Reference in New Issue
Block a user