Cover Image

On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Yin, Bo, Li, Qi, Wang, Xinchao
Format:	Preprint
Published:	2026
Subjects:	Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Refinement Provenance Inference: Detecting LLM-Refined Training Prompts from Model Behavior
by: Yin, Bo, et al.
Published: (2026)

Every Step Counts: Decoding Trajectories as Authorship Fingerprints of dLLMs
by: Li, Qi, et al.
Published: (2025)

Detecting Silent Failures in Multi-Agentic AI Trajectories
by: Pathak, Divya, et al.
Published: (2025)

AgenticEval: Toward Agentic and Self-Evolving Safety Evaluation of Large Language Models
by: Wang, Yixu, et al.
Published: (2025)

Multilingual Safety Alignment via Self-Distillation
by: Qin, Ruiyang, et al.
Published: (2026)

AT$^2$PO: Agentic Turn-based Policy Optimization via Tree Search
by: Zong, Zefang, et al.
Published: (2026)

Agentic Policy Optimization via Instruction-Policy Co-Evolution
by: Zhou, Han, et al.
Published: (2025)

From Fact Overwriting to Knowledge Evolution: Causal Editing via On-Policy Self-Distillation
by: Li, Shuaike, et al.
Published: (2026)

Discrete Diffusion in Large Language and Multimodal Models: A Survey
by: Yu, Runpeng, et al.
Published: (2025)

Multi-Level Collaboration in Model Merging
by: Li, Qi, et al.
Published: (2025)

SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
by: Li, Hao, et al.
Published: (2026)

Reasoned Safety Alignment: Ensuring Jailbreak Defense via Answer-Then-Check
by: Cao, Chentao, et al.
Published: (2025)

AgentAlign: Navigating Safety Alignment in the Shift from Informative to Agentic Large Language Models
by: Zhang, Jinchuan, et al.
Published: (2025)

TRACE: Distilling Where It Matters via Token-Routed Self On-Policy Alignment
by: Wang, Jiaxuan, et al.
Published: (2026)

Constrained Language Model Policy Optimization via Risk-aware Stepwise Alignment
by: Zhang, Lijun, et al.
Published: (2025)

Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self-Evolution
by: Qiu, Jiahao, et al.
Published: (2025)

Hierarchy-of-Groups Policy Optimization for Long-Horizon Agentic Tasks
by: He, Shuo, et al.
Published: (2026)

Position: Safety and Fairness in Agentic AI Depend on Interaction Topology, Not on Model Scale or Alignment
by: Bajaj, Tanav Singh, et al.
Published: (2026)

StructBreak: Structural Cognitive Overload-Induced Safety Failures in MLLMs
by: Luo, Yang, et al.
Published: (2026)

THINKSAFE: Self-Generated Safety Alignment for Reasoning Models
by: Lee, Seanie, et al.
Published: (2026)

Vid-SME: Membership Inference Attacks against Large Video Understanding Models
by: Li, Qi, et al.
Published: (2025)

Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment
by: Li, Jiajia, et al.
Published: (2026)

Self-Guided Defense: Adaptive Safety Alignment for Reasoning Models via Synthesized Guidelines
by: Wang, Yuhang, et al.
Published: (2025)

Demystifying the Lifecycle of Failures in Platform-Orchestrated Agentic Workflows
by: Ma, Xuyan, et al.
Published: (2025)

STAR-S: Improving Safety Alignment through Self-Taught Reasoning on Safety Rules
by: Wu, Di, et al.
Published: (2026)

HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
by: Jiang, Dongming, et al.
Published: (2026)

AI Alignment Strategies from a Risk Perspective: Independent Safety Mechanisms or Shared Failures?
by: Dung, Leonard, et al.
Published: (2025)

GraphBridge: Towards Arbitrary Transfer Learning in GNNs
by: Ju, Li, et al.
Published: (2025)

Catastrophic Failure of LLM Unlearning via Quantization
by: Zhang, Zhiwei, et al.
Published: (2024)

BAPO: Boundary-Aware Policy Optimization for Reliable Agentic Search
by: Liu, Shiyu, et al.
Published: (2026)

Solvita: Enhancing Large Language Models for Competitive Programming via Agentic Evolution
by: Li, Han, et al.
Published: (2026)

Safety Alignment via Constrained Knowledge Unlearning
by: Shi, Zesheng, et al.
Published: (2025)

Any-Depth Alignment: Unlocking Innate Safety Alignment of LLMs to Any-Depth
by: Zhang, Jiawei, et al.
Published: (2025)

SQL-ASTRA: Alleviating Sparse Feedback in Agentic SQL via Column-Set Matching and Trajectory Aggregation
by: Li, Long, et al.
Published: (2026)

Beyond Individual Intelligence: Surveying Collaboration, Failure Attribution, and Self-Evolution in LLM-based Multi-Agent Systems
by: Qi, Shihao, et al.
Published: (2026)

What Matters For Safety Alignment?
by: Li, Xing, et al.
Published: (2026)

Light Alignment Improves LLM Safety via Model Self-Reflection with a Single Neuron
by: Shen, Sicheng, et al.
Published: (2026)

Agent Safety Alignment via Reinforcement Learning
by: Sha, Zeyang, et al.
Published: (2025)

StraTA: Incentivizing Agentic Reinforcement Learning with Strategic Trajectory Abstraction
by: Xue, Xiangyuan, et al.
Published: (2026)

Policy of Thoughts: Scaling LLM Reasoning via Test-time Policy Evolution
by: Jiao, Zhengbo, et al.
Published: (2026)