Safety mid-training lowers LLM attack success before post-training
Authors report a 72% to 81% relative attack-success drop versus Qwen3 1.7B, 4B, and 8B bases.
Stories per day · 30 days
1 RSS
7 stories, newest first · one card per story
Stories tagged #attack-success-rate
Study finds LLM jailbreak graders collapse the compliance class
Across eight graders, compliance rates on two 900-response corpora differed by a factor of 22 to 25.
ASCENT: First-Order Optimal Fine-Tuning with Recalibration for Safety--Utility Co-Enhancement
ASCENT fine-tunes LLMs with periodic safety calibration, raising utility up to 20.3% and cutting attack success up to 35.5%.
Threat-Preserving Representation Sensitivity in Agent-Security Benchmarks
Agent-security benchmark ASR scores shift up to 13 percentage points when tool names change, showing single-representation robustness scores may not generalize.
EvoRiskBench: An Evolving Benchmark for Runtime Security Risks in Workspace Agents
EvoRiskBench releases 450 adversarial workspace-agent tasks; worst configuration (Codex plus DeepSeek-V4-Pro-0813) reached 68.44% attack success rate.
TRACE: Trajectory Return Attribution and Contrastive Erasure for Multi-Turn Safety
TRACE cut multi-turn attack success to the lowest rate in all 35 model-attack pairs with at most a 1.23-point utility drop.
CASCADE Against Jailbreaks: Combination Across Stages with Controlled Attack-Defense Evaluation
A controlled study finds layered jailbreak defenses beat any single defense across 19 attacks.