Imagen de Portada

Understanding and Mitigating Dataset Corruption in LLM Steering

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Anderson, Cullen, Oozeer, Narmeen, Namjoo, Foad, Ogasawara, Remy, Abdullah, Amirali, Phillips, Jeff M.
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Artificial Intelligence Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Beyond Linear Steering: Unified Multi-Attribute Control for Language Models
por: Oozeer, Narmeen, et al.
Publicado: (2025)

Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators
por: Roytburg, Dani, et al.
Publicado: (2025)

Are LLM Evaluators Really Narcissists? Sanity Checking Self-Preference Evaluations
por: Roytburg, Dani, et al.
Publicado: (2026)

Approximating Human Preferences Using a Multi-Judge Learned System
por: Sprejer, Eitán, et al.
Publicado: (2025)

Bilinear Convolution Decomposition for Causal RL Interpretability
por: Oozeer, Narmeen, et al.
Publicado: (2024)

Efficient and Stable Multi-Dimensional Kolmogorov-Smirnov Distance
por: Jacobs, Peter Matthew, et al.
Publicado: (2025)

Steer Like the LLM: Activation Steering that Mimics Prompting
por: Heyman, Geert, et al.
Publicado: (2026)

DreamReader: An Interpretability Toolkit for Text-to-Image Models
por: Prakash, Nirmalendu, et al.
Publicado: (2026)

Mitigating Overthinking in Large Reasoning Models via Manifold Steering
por: Huang, Yao, et al.
Publicado: (2025)

Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs
por: Siddique, Zara, et al.
Publicado: (2025)

Steer LLM Latents for Hallucination Detection
por: Park, Seongheon, et al.
Publicado: (2025)

Spectral Superposition: A Theory of Feature Geometry
por: Ivanov, Georgi, et al.
Publicado: (2026)

Understanding and Mitigating Bias Inheritance in LLM-based Data Augmentation on Downstream Tasks
por: Li, Miaomiao, et al.
Publicado: (2025)

Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Time
por: Zhang, Zhenyu, et al.
Publicado: (2025)

Brain-Grounded Axes for Reading and Steering LLM States
por: Andric, Sandro
Publicado: (2025)

Distribution-Aware Feature Selection for SAEs
por: Oozeer, Narmeen, et al.
Publicado: (2025)

SAEMark: Steering Personalized Multilingual LLM Watermarks with Sparse Autoencoders
por: Yu, Zhuohao, et al.
Publicado: (2025)

That's Deprecated! Understanding, Detecting, and Steering Knowledge Conflicts in Language Models for Code Generation
por: Bae, Jaesung, et al.
Publicado: (2025)

Understanding Unreliability of Steering Vectors in Language Models: Geometric Predictors and the Limits of Linear Approximations
por: Braun, Joschka
Publicado: (2026)

HyperSteer: Activation Steering at Scale with Hypernetworks
por: Sun, Jiuding, et al.
Publicado: (2025)

Understanding and Mitigating the Uncertainty in Zero-Shot Translation
por: Wang, Wenxuan, et al.
Publicado: (2022)

Understanding and Mitigating Tokenization Bias in Language Models
por: Phan, Buu, et al.
Publicado: (2024)

Compositional Steering of Large Language Models with Steering Tokens
por: Radevski, Gorjan, et al.
Publicado: (2026)

Mitigating LLM Hallucinations via Conformal Abstention
por: Yadkori, Yasin Abbasi, et al.
Publicado: (2024)

On Mitigating Code LLM Hallucinations with API Documentation
por: Jain, Nihal, et al.
Publicado: (2024)

FairFlow: Mitigating Dataset Biases through Undecided Learning
por: Cheng, Jiali, et al.
Publicado: (2025)

Directional Attractors in LLM Reasoning: How Similarity Retrieval Steers Iterative Summarization Based Reasoning
por: Tekin, Cagatay, et al.
Publicado: (2025)

Sycophancy as compositions of Atomic Psychometric Traits
por: Jain, Shreyans, et al.
Publicado: (2025)

Steering LVLMs via Sparse Autoencoder for Hallucination Mitigation
por: Hua, Zhenglin, et al.
Publicado: (2025)

Understanding LLM Embeddings for Regression
por: Tang, Eric, et al.
Publicado: (2024)

Quantifying and Mitigating Self-Preference Bias of LLM Judges
por: Yang, Jinming, et al.
Publicado: (2026)

SteeringSafety: A Systematic Safety Evaluation Framework of Representation Steering in LLMs
por: Siu, Vincent, et al.
Publicado: (2025)

What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal
por: Cheng, Stephen, et al.
Publicado: (2026)

Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMs
por: Han, Pengrui, et al.
Publicado: (2026)

Activation Space Interventions Can Be Transferred Between Large Language Models
por: Oozeer, Narmeen, et al.
Publicado: (2025)

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

Understanding Dataset Difficulty with $\mathcal{V}$-Usable Information
por: Ethayarajh, Kawin, et al.
Publicado: (2021)

Optimal Brain Iterative Merging: Mitigating Interference in LLM Merging
por: Wang, Zhixiang, et al.
Publicado: (2025)

COLD-Steer: Steering Large Language Models via In-Context One-step Learning Dynamics
por: Sharma, Kartik, et al.
Publicado: (2026)

LLM Unlearning Without an Expert Curated Dataset
por: Zhu, Xiaoyuan, et al.
Publicado: (2025)