AgentDrift Benchmark

What is AgentDrift?

Step-labeled corpus of 12,536 synthetic LLM agent tool-call trajectories in five domains, labeled per step for benign, injection point, hijacked, or failed…

Released

2026-09-07

Evaluates

Agents, General AI, Language & Knowledge

Openness

unknown

Importer

Claire Radar

Review status

ai-reviewed

Reported scores

0

Source provenance

• Original evidence https://arxiv.org/abs/2609.06972

AgentDrift paper and code

• Paper https://arxiv.org/abs/2609.06972
• Code https://github.com/Asif-0209/AgentDrift

Which sources cite AgentDrift?

• AgentDrift

Related Agents, General AI, Language & Knowledge benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• 3DCodeBench
• 5-Dialects-BN
• A Large-Scale Dataset and Benchmark
• A3S-Bench
• AARRI-Bench
• ABC-Bench
• ActBench