Cover Image

Beyond Attack Success Rate: Temporal Logit Observability for LLM Safety Failures

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Park, Junyoung, Park, Sunghwan, Ju, Seongyong, Lee, Jaewoo
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

Persona Attack: Incremental Memory Injection Jailbreak Attack against Large Language Models
by: Park, Junyoung, et al.
Published: (2026)

ASGuard: Activation-Scaling Guard to Mitigate Targeted Jailbreaking Attack
by: Park, Yein, et al.
Published: (2025)

CyBiasBench: Benchmarking Bias in LLM Agents for Cyber-Attack Scenarios
by: Lim, Taein, et al.
Published: (2026)

Hierarchical Vision Language Action Model Using Success and Failure Demonstrations
by: Park, Jeongeun, et al.
Published: (2025)

Expanding Search Space with Diverse Prompting Agents: An Efficient Sampling Approach for LLM Mathematical Reasoning
by: Lee, Gisang, et al.
Published: (2024)

Does Time Have Its Place? Temporal Heads: Where Language Models Recall Time-specific Information
by: Park, Yein, et al.
Published: (2025)

EP-HDC: Hyperdimensional Computing with Encrypted Parameters for High-Throughput Privacy-Preserving Inference
by: Park, Jaewoo, et al.
Published: (2025)

Adaptive Replay Buffer for Offline-to-Online Reinforcement Learning
by: Song, Chihyeon, et al.
Published: (2025)

MolDeTox: Evaluating Language Model's Stepwise Fragment Editing for Molecular Detoxification
by: Park, Jueon, et al.
Published: (2026)

Bridging Temporal and Textual Modalities: A Multimodal Framework for Automated Cloud Failure Root Cause Analysis
by: Park, Gijun
Published: (2026)

Coherent Without Grounding, Grounded Without Success: Observability and Epistemic Failure
by: Sartori, Camilo Chacón
Published: (2026)

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

QUOKA: Query-Oriented KV Selection For Efficient LLM Prefill
by: Jones, Dalton, et al.
Published: (2026)

Fast Forward: Accelerating LLM Prefill with Predictive FFN Sparsity
by: Gautam, Aayush, et al.
Published: (2026)

Thinking Sparks!: Emergent Attention Heads in Reasoning Models During Post Training
by: Park, Yein, et al.
Published: (2025)

KeyDiff: Key Similarity-Based KV Cache Eviction for Long-Context LLM Inference in Resource-Constrained Environments
by: Park, Junyoung, et al.
Published: (2025)

Uncovering Logit Suppression Vulnerabilities in LLM Safety Alignment
by: Li, Yuxi, et al.
Published: (2024)

Recursive Speculative Decoding: Accelerating LLM Inference via Sampling Without Replacement
by: Jeon, Wonseok, et al.
Published: (2024)

ChartCap: Mitigating Hallucination of Dense Chart Captioning
by: Lim, Junyoung, et al.
Published: (2025)

Tracing Mathematical Proficiency Through Problem-Solving Processes
by: Park, Jungyang, et al.
Published: (2025)

Face Reconstruction Transfer Attack as Out-of-Distribution Generalization
by: Jung, Yoon Gyo, et al.
Published: (2024)

Guided Trajectory Generation with Diffusion Models for Offline Model-based Optimization
by: Yun, Taeyoung, et al.
Published: (2024)

GTA: Generative Trajectory Augmentation with Guidance for Offline Reinforcement Learning
by: Lee, Jaewoo, et al.
Published: (2024)

CoTox: Chain-of-Thought-Based Molecular Toxicity Reasoning and Prediction
by: Park, Jueon, et al.
Published: (2025)

Distillability of LLM Security Logic: Predicting Attack Success Rate of Outline Filling Attack via Ranking Regression
by: Zhang, Tianyu, et al.
Published: (2025)

ChroKnowledge: Unveiling Chronological Knowledge of Language Models in Multiple Domains
by: Park, Yein, et al.
Published: (2024)

Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR
by: Lee, Chanuk, et al.
Published: (2026)

Single-Configuration Attack Success Rate Is Not Enough: Jailbreak Evaluations Should Report Distributional Attack Success
by: Maple, Carsten, et al.
Published: (2026)

Steering Language Models Before They Speak: Logit-Level Interventions
by: An, Hyeseon, et al.
Published: (2026)

ToxReason: A Benchmark for Mechanistic Chemical Toxicity Reasoning via Adverse Outcome Pathway
by: Park, Jueon, et al.
Published: (2026)

Early Diagnosis of Wasted Computation in Multi-Agent LLM Systems via Failure-Aware Observability
by: Li, Xianyou, et al.
Published: (2026)

PrismQuant: Rate-Distortion-Optimal Vector Quantization for Gaussian-Mixture Sources
by: Park, Bumsu, et al.
Published: (2026)

Automated Kernel Discovery Towards Understanding High-dimensional Bayesian Optimization
by: Yun, Taeyoung, et al.
Published: (2026)

Diffusion Fine-Tuning via Reparameterized Policy Gradient of the Soft Q-Function
by: Kang, Hyeongyu, et al.
Published: (2025)

Beyond Line-Level Filtering for the Pretraining Corpora of LLMs
by: Park, Chanwoo, et al.
Published: (2025)

Beyond Attack Success Rate: A Multi-Metric Evaluation of Adversarial Transferability in Medical Imaging Models
by: Curl, Emily, et al.
Published: (2026)

Reasoning Structure Matters for Safety Alignment of Reasoning Models
by: In, Yeonjun, et al.
Published: (2026)

Peak-Controlled Logits Poisoning Attack in Federated Distillation
by: Tang, Yuhan, et al.
Published: (2024)

Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games
by: Park, Dongmin, et al.
Published: (2025)

From Ambiguity to Accuracy: The Transformative Effect of Coreference Resolution on Retrieval-Augmented Generation systems
by: Jang, Youngjoon, et al.
Published: (2025)