Researchagentsllmbackdoorbenchmarks
LLM Agents Expose Persistent Backdoor Vulnerabilities
9.0
Researchers present BackdoorAgent, a modular, stage-aware framework and benchmark for analyzing backdoor threats in LLM agents, submitted Jan. 8, 2026. They structure attacks into planning, memory, and tool-use stages and evaluate across four agent applications (Agent QA, Code, Web, Drive), finding trigger persistence rates of 43.58% (planning), 77.97% (memory), and 60.28% (tool-stage) on a GPT-based backbone. The authors release code and benchmark on GitHub.
Key Points
- 1Demonstrate triggers persist across agent stages, with 43.58% planning, 77.97% memory, 60.28% tool-stage
- 2Show cross-stage propagation enlarges agent attack surface, exposing multi-step workflows to sustained manipulation
- 3Advise practitioners to monitor planning, memory, and tool modules, and test end-to-end trigger persistence
Scoring Rationale
Strong novelty and actionable benchmark; limited by single preprint source and narrow GPT-based evaluation scope.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
High-Value Overnight OrdersEasyDelivered International ShipmentsMediumOn-Time Delivery Rate by CarrierHard
250 free problems · No credit card
See all Logistics & Shipping problems