InfoResearchPreprintLLM-specific
Evolution Under Fire: Evolving Agent Safety Harnesses with Adaptive Injection Stressors
- Published
- Record updated
Summary
Evolution Under Fire is a staged framework that co-evolves executable safety harnesses for tool-using LLM agents together with adaptive injection stressors, keeping the task model, benchmark, tools and evaluator fixed. On AgentDyn with GPT-4o-mini, an evolved harness reduces attack success from 50.4% to 3.0% while improving utility. On Agent-SafetyBench with Qwen2.5-7B-Instruct, attack success falls from 82.6% to 62.0%, and harnesses transfer to Qwen3.5-35B-A3B, cutting it from 32.1% to 0.9%.
Related items
- InfoClaude-maker Anthropic poised for $1.4tn stock market debutSame vendor · BBC Technology
- InfoRemember the frontier AI pause? Chip companies don’t eitherSame vendor · The Verge (AI)
- InfoThe seven Anthropic founders who are about to become multi-billionairesSame vendor · BBC Technology
- MediumCVE-2026-132122: agentic-radar untrusted .env loading leaks OpenAI API key via scan commandSame vendor · NVD/CVE Database
- InfoThe Silent ConsensusSame vendor · OpenAI Blog