Imagen de Portada

Perturbation Probing: A Two-Pass-per-Prompt Diagnostic for FFN Behavioral Circuits in Aligned LLMs

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Liu, Hongliang, Li, Tung-Ling, Wu, Yuhao
Formato:	Preprint
Publicado:	2026
Materias:	Computation and Language Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Logit-Gap Steering: A Forward-Pass Diagnostic for Alignment Robustness
por: Li, Tung-Ling, et al.
Publicado: (2025)

AdvJudge-Zero: Binary Decision Flips in LLM-as-a-Judge via Adversarial Control Tokens
por: Li, Tung-Ling, et al.
Publicado: (2025)

Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows
por: Xu, Wanghan, et al.
Publicado: (2025)

Flip-Flop Consistency: Unsupervised Training for Robustness to Prompt Perturbations in LLMs
por: Hejabi, Parsa, et al.
Publicado: (2025)

FFN-SkipLLM: A Hidden Gem for Autoregressive Decoding with Adaptive Feed Forward Skipping
por: Jaiswal, Ajay, et al.
Publicado: (2024)

BlockFFN: Towards End-Side Acceleration-Friendly Mixture-of-Experts with Chunk-Level Activation Sparsity
por: Song, Chenyang, et al.
Publicado: (2025)

Aligning Multiple Knowledge Graphs in a Single Pass
por: Yang, Yaming, et al.
Publicado: (2024)

Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates
por: Lyu, Kaifeng, et al.
Publicado: (2024)

Circuit Breaking: Removing Model Behaviors with Targeted Ablation
por: Li, Maximilian, et al.
Publicado: (2023)

Prompt Perturbation Consistency Learning for Robust Language Models
por: Qiang, Yao, et al.
Publicado: (2024)

Aligning (Medical) LLMs for (Counterfactual) Fairness
por: Poulain, Raphael, et al.
Publicado: (2024)

H3Fusion: Helpful, Harmless, Honest Fusion of Aligned LLMs
por: Tekin, Selim Furkan, et al.
Publicado: (2024)

Patch-Prompt Aligned Bayesian Prompt Tuning for Vision-Language Models
por: Liu, Xinyang, et al.
Publicado: (2023)

MATH-Perturb: Benchmarking LLMs' Math Reasoning Abilities against Hard Perturbations
por: Huang, Kaixuan, et al.
Publicado: (2025)

Everybody Prune Now: Structured Pruning of LLMs with only Forward Passes
por: Kolawole, Steven, et al.
Publicado: (2024)

The Metacognitive Probe: Five Behavioural Calibration Diagnostics for LLMs
por: Oliveira, Rafael C. T.
Publicado: (2026)

Benchmarking and Understanding Compositional Relational Reasoning of LLMs
por: Ni, Ruikang, et al.
Publicado: (2024)

Vaccine: Perturbation-aware Alignment for Large Language Models against Harmful Fine-tuning Attack
por: Huang, Tiansheng, et al.
Publicado: (2024)

Are the Values of LLMs Structurally Aligned with Humans? A Causal Perspective
por: Kang, Yipeng, et al.
Publicado: (2024)

CALF: Aligning LLMs for Time Series Forecasting via Cross-modal Fine-Tuning
por: Liu, Peiyuan, et al.
Publicado: (2024)

Can we Soft Prompt LLMs for Graph Learning Tasks?
por: Liu, Zheyuan, et al.
Publicado: (2024)

Selective Prompting Tuning for Personalized Conversations with LLMs
por: Huang, Qiushi, et al.
Publicado: (2024)

LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression
por: Jiang, Huiqiang, et al.
Publicado: (2023)

Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration
por: Huang, Langlin, et al.
Publicado: (2026)

Large Language Model Unlearning via Embedding-Corrupted Prompts
por: Liu, Chris Yuhao, et al.
Publicado: (2024)

Aligning LLMs by Predicting Preferences from User Writing Samples
por: Aroca-Ouellette, Stéphane, et al.
Publicado: (2025)

ALIEN: Aligned Entropy Head for Improving Uncertainty Estimation of LLMs
por: Zabolotnyi, Artem, et al.
Publicado: (2025)

Dialectical Behavior Therapy Approach to LLM Prompting
por: Vitman, Oxana, et al.
Publicado: (2024)

Can Prompts Rewind Time for LLMs? Evaluating the Effectiveness of Prompted Knowledge Cutoffs
por: Gao, Xin, et al.
Publicado: (2025)

Improving Preference Extraction In LLMs By Identifying Latent Knowledge Through Classifying Probes
por: Maiya, Sharan, et al.
Publicado: (2025)

Probing the Decision Boundaries of In-context Learning in Large Language Models
por: Zhao, Siyan, et al.
Publicado: (2024)

Dynamic Prompt Fusion for Multi-Task and Cross-Domain Adaptation in LLMs
por: Hu, Xin, et al.
Publicado: (2025)

Aligning the Spectrum: Hybrid Graph Pre-training and Prompt Tuning across Homophily and Heterophily
por: Luo, Haitong, et al.
Publicado: (2025)

Improving Quantized Model Performance in Qualitative Analysis with Multi-Pass Prompt Verification
por: Adeseye, Aisvarya, et al.
Publicado: (2026)

Simplify-This: A Comparative Analysis of Prompt-Based and Fine-Tuned LLMs
por: Cohen, Eilam, et al.
Publicado: (2026)

Probing to Refine: Reinforcement Distillation of LLMs via Explanatory Inversion
por: Tan, Zhen, et al.
Publicado: (2026)

The Alignment Tax: Response Homogenization in Aligned LLMs and Its Implications for Uncertainty Estimation
por: Liu, Mingyi
Publicado: (2026)

Looking for the Inner Music: Probing LLMs' Understanding of Literary Style
por: Hicke, Rebecca M. M., et al.
Publicado: (2025)

StreetMath: Study of LLMs' Approximation Behaviors
por: Tseng, Chiung-Yi, et al.
Publicado: (2025)

More Than a Score: Probing the Impact of Prompt Specificity on LLM Code Generation
por: Zi, Yangtian, et al.
Publicado: (2025)