Copertina

Weight Updates as Activation Shifts: A Principled Framework for Steering

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Adila, Dyah, Cooper, John, Yun, Alexander, Trost, Avi, Sala, Frederic
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Personalize Your LLM: Fake it then Align it
di: Zhang, Yijing, et al.
Pubblicazione: (2025)

Zero-Shot Robustification of Zero-Shot Models
di: Adila, Dyah, et al.
Pubblicazione: (2023)

Is Free Self-Alignment Possible?
di: Adila, Dyah, et al.
Pubblicazione: (2024)

Multimodal Data Curation via Object Detection and Filter Ensembles
di: Huang, Tzu-Heng, et al.
Pubblicazione: (2024)

Grow, Don't Overwrite: Fine-tuning Without Forgetting
di: Adila, Dyah, et al.
Pubblicazione: (2026)

Discovering Bias in Latent Space: An Unsupervised Debiasing Approach
di: Adila, Dyah, et al.
Pubblicazione: (2024)

Evaluating Sample Utility for Efficient Data Selection by Mimicking Model Weights
di: Huang, Tzu-Heng, et al.
Pubblicazione: (2025)

Weak-to-Strong Generalization Through the Data-Centric Lens
di: Shin, Changho, et al.
Pubblicazione: (2024)

CrEst: Credibility Estimation for Contexts in LLMs via Weak Supervision
di: Adila, Dyah, et al.
Pubblicazione: (2025)

Expressivity-Efficiency Tradeoffs for Hybrid Sequence Models
di: Cooper, John, et al.
Pubblicazione: (2026)

R&B: Domain Regrouping and Data Mixture Balancing for Efficient Foundation Model Training
di: Ge, Albert, et al.
Pubblicazione: (2025)

TARDIS: Mitigating Temporal Misalignment via Representation Steering
di: Shin, Changho, et al.
Pubblicazione: (2025)

Test-Time Scaling Makes Overtraining Compute-Optimal
di: Roberts, Nicholas, et al.
Pubblicazione: (2026)

Guiding Giants: Lightweight Controllers for Weighted Activation Steering in LLMs
di: Hegazy, Amr, et al.
Pubblicazione: (2025)

Learning to Generate Instruction Tuning Datasets for Zero-Shot Task Adaptation
di: Nayak, Nihal V., et al.
Pubblicazione: (2024)

Steering Language Models With Activation Engineering
di: Turner, Alexander Matt, et al.
Pubblicazione: (2023)

Quantifying Structure in CLIP Embeddings: A Statistical Framework for Concept Interpretation
di: Zhao, Jitian, et al.
Pubblicazione: (2025)

Towards Steering without Sacrifice: Principled Training of Steering Vectors for Prompt-only Interventions
di: Bao, Yuntai, et al.
Pubblicazione: (2026)

Activation Steering with a Feedback Controller
di: Nguyen, Dung V., et al.
Pubblicazione: (2025)

Dynamically Scaled Activation Steering
di: Ferrando, Alex, et al.
Pubblicazione: (2025)

Curvature-Guided LoRA: Steering in the pretrained NTK subspace
di: Zheng, Frédéric, et al.
Pubblicazione: (2026)

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

Steering Language Models with Weight Arithmetic
di: Fierro, Constanza, et al.
Pubblicazione: (2025)

Label Shift Estimation With Incremental Prior Update
di: Zhang, Yunrui, et al.
Pubblicazione: (2026)

Linearly Constrained Weights: Reducing Activation Shift for Faster Training of Neural Networks
di: Kutsuna, Takuro
Pubblicazione: (2024)

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

MAGDiff: Covariate Data Set Shift Detection via Activation Graphs of Deep Neural Networks
di: Arnal, Charles, et al.
Pubblicazione: (2023)

Minimizing Collateral Damage in Activation Steering
di: Nguyen, Tam, et al.
Pubblicazione: (2026)

Steered LLM Activations are Non-Surjective
di: Mishra, Aayush, et al.
Pubblicazione: (2026)

Activation Steering for Chain-of-Thought Compression
di: Azizi, Seyedarmin, et al.
Pubblicazione: (2025)

Towards Representation Learning for Weighting Problems in Design-Based Causal Inference
di: Clivio, Oscar, et al.
Pubblicazione: (2024)

Depth-Wise Activation Steering for Honest Language Models
di: Góral, Gracjan, et al.
Pubblicazione: (2025)

Steering Pretrained Drafters during Speculative Decoding
di: Berdoz, Frédéric, et al.
Pubblicazione: (2025)

AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint
di: Sheng, Leheng, et al.
Pubblicazione: (2025)

Steering Llama 2 via Contrastive Activation Addition
di: Panickssery, Nina, et al.
Pubblicazione: (2023)

Beyond Steering Vector: Flow-based Activation Steering for Inference-Time Intervention
di: Jin, Zehao, et al.
Pubblicazione: (2026)

Command-V: Pasting LLM Behaviors via Activation Profiles
di: Wang, Barry, et al.
Pubblicazione: (2025)

Steering Code LLMs with Activation Directions for Language and Library Control
di: Rahman, Md Mahbubur, et al.
Pubblicazione: (2026)

Global Evolutionary Steering: Refining Activation Steering Control via Cross-Layer Consistency
di: Jiang, Xinyan, et al.
Pubblicazione: (2026)

Steering Without Side Effects: Improving Post-Deployment Control of Language Models
di: Stickland, Asa Cooper, et al.
Pubblicazione: (2024)