Image de couverture de livre

REFLECTOR: Internalizing Step-wise Reflection against Indirect Jailbreak

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Ma, Jiachen, Zhang, Jiawen, Li, Xiangtian, Zou, Bo, Lu, Chaochao, Yang, Chao
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Native Reasoning Models: Training Language Models to Reason on Unverifiable Data
par: Wang, Yuanfu, et autres
Publié: (2026)

SafeCoT: Improving VLM Safety with Minimal Reasoning
par: Ma, Jiachen, et autres
Publié: (2025)

Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMs
par: Lai, Xin, et autres
Publié: (2024)

Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking
par: He, Zhida, et autres
Publié: (2026)

RePO: Replay-Enhanced Policy Optimization
par: Li, Siheng, et autres
Publié: (2025)

AudioJailbreak: Jailbreak Attacks against End-to-End Large Audio-Language Models
par: Chen, Guangke, et autres
Publié: (2025)

CauScale: Neural Causal Discovery at Scale
par: Peng, Bo, et autres
Publié: (2026)

ALERT: Zero-shot LLM Jailbreak Detection via Internal Discrepancy Amplification
par: Lin, Xiao, et autres
Publié: (2026)

Structure-based RNA Design by Step-wise Optimization of Latent Diffusion Model
par: Si, Qi, et autres
Publié: (2026)

CauScientist: Teaching LLMs to Respect Data for Causal Discovery
par: Peng, Bo, et autres
Publié: (2026)

SSPO: Self-traced Step-wise Preference Optimization for Process Supervision and Reasoning Compression
par: Xu, Yuyang, et autres
Publié: (2025)

Towards AI-$45^{\circ}$ Law: A Roadmap to Trustworthy AGI
par: Yang, Chao, et autres
Publié: (2024)

Towards Lightweight Time Series Forecasting: a Patch-wise Transformer with Weak Data Enriching
par: Wang, Meng, et autres
Publié: (2025)

Omanic: Towards Step-wise Evaluation of Multi-hop Reasoning in Large Language Models
par: Gu, Xiaojie, et autres
Publié: (2026)

Reasoned Safety Alignment: Ensuring Jailbreak Defense via Answer-Then-Check
par: Cao, Chentao, et autres
Publié: (2025)

Understanding and Preserving Safety in Fine-Tuned LLMs
par: Zhang, Jiawen, et autres
Publié: (2026)

Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning
par: Deng, Yihe, et autres
Publié: (2025)

Step-wise Adaptive Integration of Supervised Fine-tuning and Reinforcement Learning for Task-Specific LLMs
par: Chen, Jack, et autres
Publié: (2025)

Augmented Lagrangian Multiplier Network for State-wise Safety in Reinforcement Learning
par: Zhang, Jiaming, et autres
Publié: (2026)

Self-Guided Process Reward Optimization with Redefined Step-wise Advantage for Process Reinforcement Learning
par: Fei, Wu, et autres
Publié: (2025)

Self-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories
par: Zhang, Dongcheng, et autres
Publié: (2026)

Potent but Stealthy: Rethink Profile Pollution against Sequential Recommendation via Bi-level Constrained Reinforcement Paradigm
par: Su, Jiajie, et autres
Publié: (2025)

Epistemic Traps: Rational Misalignment Driven by Model Misspecification
par: Xu, Xingcheng, et autres
Publié: (2026)

From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step
par: Deng, Yuntian, et autres
Publié: (2024)

BC Protocol: Structured Dual-Expert Dialogue for Eliciting High-Quality Chain-of-Thought Post-Training Data
par: Zou, Bo, et autres
Publié: (2026)

QUIET: A Multi-Blank Cascaded Story Cloze Benchmark for LLM Creative Generation Capability
par: Zou, Bo, et autres
Publié: (2026)

Calibrated Surprise: An Information-Theoretic Account of Creative Quality
par: Zou, Bo, et autres
Publié: (2026)

Jailbreaking Black Box Large Language Models in Twenty Queries
par: Chao, Patrick, et autres
Publié: (2023)

HSF: Defending against Jailbreak Attacks with Hidden State Filtering
par: Qian, Cheng, et autres
Publié: (2024)

A Novel Multimodal RUL Framework for Remaining Useful Life Estimation with Layer-wise Explanations
par: Razzaq, Waleed, et autres
Publié: (2025)

Can Post-Training Transform LLMs into Causal Reasoners?
par: Chen, Junqi, et autres
Publié: (2026)

Adversarial Curriculum Graph Contrastive Learning with Pair-wise Augmentation
par: Zhao, Xinjian, et autres
Publié: (2024)

The Struggle Between Continuation and Refusal: A Mechanistic Analysis of the Continuation-Triggered Jailbreak in LLMs
par: Deng, Yonghong, et autres
Publié: (2026)

LINA: Learning INterventions Adaptively for Physical Alignment and Generalization in Diffusion Models
par: Yu, Shu, et autres
Publié: (2025)

Equivariant Evidential Deep Learning for Interatomic Potentials
par: Wang, Zhongyao, et autres
Publié: (2026)

Dynamic Adversarial Reinforcement Learning for Robust Multimodal Large Language Models
par: Bao, Yicheng, et autres
Publié: (2026)

Efficient ANN-Guided Distillation: Aligning Rate-based Features of Spiking Neural Networks through Hybrid Block-wise Replacement
par: Yang, Shu, et autres
Publié: (2025)

UDora: A Unified Red Teaming Framework against LLM Agents by Dynamically Hijacking Their Own Reasoning
par: Zhang, Jiawei, et autres
Publié: (2025)

The Jailbreak Tax: How Useful are Your Jailbreak Outputs?
par: Nikolić, Kristina, et autres
Publié: (2025)

Scalable Defense against In-the-wild Jailbreaking Attacks with Safety Context Retrieval
par: Chen, Taiye, et autres
Publié: (2025)