TERM: Target-Free Jailbreak Defense via Entropy-Guided Residual Modulation
cs.CR, cs.CL
Submitted: 2026-11-10
Updated: 2026-11-10
Terminology
Sources
- Detecting Fluent Optimization-Based Adversarial Prompts via Sequential Entropy Changes
- Training Verifiers to Solve Math Word Problems
- Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators
- The Llama 3 Herd of Models
- Measuring Massive Multitask Language Understanding
- Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations
- RobustKV: Defending Large Language Models against Jailbreak Attacks via KV Eviction
- Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment
- HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal
- Large Language Models: A Survey
- MirrorShield: Towards Universal Defense Against Jailbreaks via Entropy-Guided Mirror Crafting
- Qwen2.5 Technical Report
- Scalable and Transferable Black-Box Jailbreaks for Language Models via Persona Modulation
- Gemma 2: Improving Open Language Models at a Practical Size
- SafeInt: Shielding Large Language Models from Jailbreak Attacks via Safety-Aware Representation Intervention
- PARDEN, Can You Repeat That? Defending against Jailbreaks via Repetition
- Universal and Transferable Adversarial Attacks on Aligned Language Models
Related papers
- From Strategy Search to Policy Learning: MCTS-Guided Adversarial Co-Evolution for LLM Safety Alignment
- Privacy Leakage from Recurrent State in LLM Inference
- Incrementally Verifiable Oracle Aided Computation
- Privilege and confidentiality in generative AI workflows
- BadMoE: Backdooring Mixture-of-Experts LLMs via Routing-Aware Triggers and Task-Underutilized Experts
- CoVault: Secure, Scalable Analytics of Personal Data