Cover Image

Steerable but Not Decodable: Function Vectors Operate Beyond the Logit Lens

Fuente: arXiv

Saved in:

Bibliographic Details
Main Author:	Nadaf, Mohammed Suhail B
Format:	Preprint
Published:	2026
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

reward-lens: A Mechanistic Interpretability Library for Reward Models
by: Nadaf, Mohammed Suhail B
Published: (2026)

Through a Steerable Lens: Magnifying Neural Network Interpretability via Phase-Based Extrapolation
by: Mahdisoltani, Farzaneh, et al.
Published: (2025)

Contextual Multinomial Logit Bandits with General Value Functions
by: Zhang, Mengxiao, et al.
Published: (2024)

Provably Efficient Reinforcement Learning with Multinomial Logit Function Approximation
by: Li, Long-Fei, et al.
Published: (2024)

Beyond Logit Adjustment: A Residual Decomposition Framework for Long-Tailed Reranking
by: Wang, Zhanliang, et al.
Published: (2026)

From Projection to Prediction: Beyond Logits for Scalable Language Models
by: Dong, Jianbing, et al.
Published: (2025)

Logits-Based Finetuning
by: Li, Jingyao, et al.
Published: (2025)

Stable and Steerable Sparse Autoencoders with Weight Regularization
by: Jedryszek, Piotr, et al.
Published: (2026)

Spectral Image Tokenizer
by: Esteves, Carlos, et al.
Published: (2024)

Variance-Adaptive Optimal Algorithm for Reinforcement Learning with Multinomial Logit Function Approximation
by: Kim, Wonyoung, et al.
Published: (2026)

GPT-2 Through the Lens of Vector Symbolic Architectures
by: Knittel, Johannes, et al.
Published: (2024)

SLED: Self Logits Evolution Decoding for Improving Factuality in Large Language Models
by: Zhang, Jianyi, et al.
Published: (2024)

Beyond Hidden-Layer Manipulation: Semantically-Aware Logit Interventions for Debiasing LLMs
by: Xia, Wei
Published: (2025)

Steerable Neural ODEs on Homogeneous Spaces
by: Andersdotter, Emma, et al.
Published: (2026)

Learning Steerable Clarification Policies with Collaborative Self-play
by: Berant, Jonathan, et al.
Published: (2025)

Clifford-Steerable Convolutional Neural Networks
by: Zhdanov, Maksim, et al.
Published: (2024)

The Implicit Bias of Logit Regularization
by: Beck, Alon, et al.
Published: (2026)

DeLTa: A Decoding Strategy based on Logit Trajectory Prediction Improves Factuality and Reasoning Ability
by: He, Yunzhen, et al.
Published: (2025)

Inferring from Logits: Exploring Best Practices for Decoding-Free Generative Candidate Selection
by: Ma, Mingyu Derek, et al.
Published: (2025)

A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs
by: Veefkind, Lars, et al.
Published: (2024)

Beyond Real Data: Synthetic Data through the Lens of Regularization
by: Shidani, Amitis, et al.
Published: (2025)

Learning Policy Representations for Steerable Behavior Synthesis
by: Li, Beiming, et al.
Published: (2026)

Generative Molecular Design with Steerable and Granular Synthesizability Control
by: Guo, Jeff, et al.
Published: (2025)

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

Improving Actor-Critic Training with Steerable Action-Value Approximation Errors
by: Tasdighi, Bahareh, et al.
Published: (2024)

DALD: Improving Logits-based Detector without Logits from Black-box LLMs
by: Zeng, Cong, et al.
Published: (2024)

Top-$nσ$: Not All Logits Are You Need
by: Tang, Chenxia, et al.
Published: (2024)

Exploiting Features and Logits in Heterogeneous Federated Learning
by: Chan, Yun-Hin, et al.
Published: (2022)

Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment
by: Zhang, Yunfan, et al.
Published: (2025)

Beyond Scaling Curves: Internal Dynamics of Neural Networks Through the NTK Lens
by: Nikolaou, Konstantin, et al.
Published: (2025)

Interpretable and Steerable Concept Bottleneck Sparse Autoencoders
by: Kulkarni, Akshay, et al.
Published: (2025)

Compositional Generalization in Autoregressive Models via Logit Composition
by: Kumar, Aakash, et al.
Published: (2026)

Online Continual Learning via Logit Adjusted Softmax
by: Huang, Zhehao, et al.
Published: (2023)

Consistency Regularization for Domain Generalization with Logit Attribution Matching
by: Gao, Han, et al.
Published: (2023)

Learning to Scale Logits for Temperature-Conditional GFlowNets
by: Kim, Minsu, et al.
Published: (2023)

Bayesian Preference Learning for Test-Time Steerable Reward Models
by: Hong, Jiwoo, et al.
Published: (2026)

Building a Scalable, Effective, and Steerable Search and Ranking Platform
by: Celikik, Marjan, et al.
Published: (2024)

Steerable Scene Generation with Post Training and Inference-Time Search
by: Pfaff, Nicholas, et al.
Published: (2025)

Logit Distance Bounds Representational Similarity
by: Nielsen, Beatrix M. G., et al.
Published: (2026)

Logit Reweighting for Topic-Focused Summarization
by: Braun, Joschka, et al.
Published: (2025)