TERM jailbreak defense modulates residuals from attention entropy
Compared with seven defenses on seven attacks, it reports stronger robustness and low over-refusal at modest inference cost.
Stories per day · 30 days
2 RSS
Relatedattention2guardrails2jailbreak2llm-defense2alignment1benchmark1dpo1harmbench1inference-time1jailbreakbench1llm-judge1llm-safety1manaca1multi-agent1portuguese1prompt-injection1residual-modulation1safety-alignment1
8 stories, newest first · one card per story
Stories tagged #over-refusal
ErosionBench scores alignment decay in enterprise multi-agent systems
The benchmark has 1,728 scenarios and 198 image attacks; a Meta-Auditor reached 98.5% recall at 2.9% false positives.
Jev LLM safety guardrail reduces harmful outputs
Jev input filtering and selective output review cut harmful LLM responses from 73.2% to 2.4% while limiting benign refusals.
Guardrail safety scores track policy labels more than detection capability
Ten arms trained on about 53k dual-labeled prompts show HarmBench-copyright scores spanning 0-100 while detection stays unchanged.
Manacá-1B-Instruct-2: Reproducible Safety Alignment for a Native Brazilian-Portuguese LLM. Negative DPO Results, a Controlled Composition Scan, and a Silent Failure Mode of the LLM Judge
Manacá-1B-Instruct-2 gained safety from iterated safety finetuning after three failed DPO runs, beating four Portuguese baselines.
PatchBench: Measuring Collateral Damage in Activation Patching
PatchBench checks whether LLM jailbreak repairs fix nearby harmful prompts without over-refusing similar benign ones.
When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety
Matched tests find DPO controls LLM safety best, while representation probes monitor cheaply and can repair fine-tuning damage.
Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal
A self-distillation safety framework tunes narrow-boundary refusals in Qwen3-8B, raising target-domain refusal to 84.75% while cutting over-refusal from 15.20% to 5.20%.