Image de couverture de livre

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Wu, Zhengxuan, Arora, Aryaman, Geiger, Atticus, Wang, Zheng, Huang, Jing, Jurafsky, Dan, Manning, Christopher D., Potts, Christopher
Format:	Preprint
Publié:	2025
Sujets:	Computation and Language Artificial Intelligence Machine Learning I.2.7
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

CausalGym: Benchmarking causal interpretability methods on linguistic tasks
par: Arora, Aryaman, et autres
Publié: (2024)

Mechanistic evaluation of Transformers and state space models
par: Arora, Aryaman, et autres
Publié: (2025)

Bayesian scaling laws for in-context learning
par: Arora, Aryaman, et autres
Publié: (2024)

Language Model Circuits Are Sparse in the Neuron Basis
par: Arora, Aryaman, et autres
Publié: (2026)

ADAG: Automatically Describing Attribution Graphs
par: Arora, Aryaman, et autres
Publié: (2026)

CRISP: Persistent Concept Unlearning via Sparse Autoencoders
par: Ashuach, Tomer, et autres
Publié: (2025)

Tversky Neural Networks: Psychologically Plausible Deep Learning with Differentiable Tversky Similarity
par: Doumbouya, Moussa Koulako Bala, et autres
Publié: (2025)

Control Reinforcement Learning: Interpretable Token-Level Steering of LLMs via Sparse Autoencoder Features
par: Cho, Seonglae, et autres
Publié: (2026)

ReFT: Representation Finetuning for Language Models
par: Wu, Zhengxuan, et autres
Publié: (2024)

CorrSteer: Generation-Time LLM Steering via Correlated Sparse Autoencoder Features
par: Cho, Seonglae, et autres
Publié: (2025)

Learning the meanings of function words from grounded language using a visual question answering model
par: Portelance, Eva, et autres
Publié: (2023)

Steer-MoE: Efficient Audio-Language Alignment with a Mixture-of-Experts Steering Module
par: Feng, Ruitao, et autres
Publié: (2025)

Merge-Bench: Resolve Merge Conflicts with Large Language Models
par: Schesch, Benedikt, et autres
Publié: (2026)

Detecting and Steering LLMs' Empathy in Action
par: Cadile, Juan P.
Publié: (2025)

ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs
par: Simhi, Adi, et autres
Publié: (2025)

PersistBench: When Should Long-Term Memories Be Forgotten by LLMs?
par: Pulipaka, Sidharth, et autres
Publié: (2026)

Descriptive Collision in Sparse Autoencoder Auto-Interpretability: When One Explanation Describes Many Features
par: McCann, Jordan F.
Publié: (2026)

RomanLens: The Role Of Latent Romanization In Multilinguality In LLMs
par: Saji, Alan, et autres
Publié: (2025)

Curveball Steering: The Right Direction To Steer Isn't Always Linear
par: Raval, Shivam, et autres
Publié: (2026)

Fine-Tuned 'Small' LLMs (Still) Significantly Outperform Zero-Shot Generative AI Models in Text Classification
par: Bucher, Martin Juan José, et autres
Publié: (2024)

EduGuardBench: A Holistic Benchmark for Evaluating the Pedagogical Fidelity and Adversarial Safety of LLMs as Simulated Teachers
par: Jiang, Yilin, et autres
Publié: (2025)

Reducing Information Overload: Because Even Security Experts Need to Blink
par: Kuehn, Philipp, et autres
Publié: (2022)

Enhancing Paraphrase Type Generation: The Impact of DPO and RLHF Evaluated with Human-Ranked Data
par: Lübbers, Christopher Lee
Publié: (2025)

ExpressivityBench: Can LLMs Communicate Implicitly?
par: Tint, Joshua, et autres
Publié: (2024)

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
par: Agrawal, Lakshya A, et autres
Publié: (2025)

Evaluating Steering Techniques using Human Similarity Judgments
par: Studdiford, Zach, et autres
Publié: (2025)

PathBench: Speech Intelligibility Benchmark for Automatic Pathological Speech Assessment
par: Halpern, Bence Mark, et autres
Publié: (2026)

LoRS: Efficient Low-Rank Adaptation for Sparse Large Language Model
par: Hu, Yuxuan, et autres
Publié: (2025)

Exploring RWKV for Sentence Embeddings: Layer-wise Analysis and Baseline Comparison for Semantic Similarity
par: Pan, Xinghan
Publié: (2025)

Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders
par: Patel, Het, et autres
Publié: (2026)

PaperAudit-Bench: Benchmarking Error Detection in Research Papers for Critical Automated Peer Review
par: Tu, Songjun, et autres
Publié: (2026)

Combining Denoising Autoencoders with Contrastive Learning to fine-tune Transformer Models
par: Lopez-Avila, Alejo, et autres
Publié: (2024)

Simple and Effective Baselines for Code Summarisation Evaluation
par: Robinson, Jade, et autres
Publié: (2025)

Steering Language Models in Multi-Token Generation: A Case Study on Tense and Aspect
par: Klerings, Alina, et autres
Publié: (2025)

AsyncTLS: Efficient Generative LLM Inference with Asynchronous Two-level Sparse Attention
par: Hu, Yuxuan, et autres
Publié: (2026)

BabyLlama-2: Ensemble-Distilled Models Consistently Outperform Teachers With Limited Data
par: Tastet, Jean-Loup, et autres
Publié: (2024)

UA-Legal-Bench: A Benchmark for Evaluating Large Language Models on Ukrainian Legal Reasoning
par: Ovcharov, Volodymyr
Publié: (2026)

RTI-Bench: A Structured Dataset for Indian Right-to-Information Decision Analysis
par: Bose, Joy
Publié: (2026)

Adaptive Focus Memory for Language Models
par: Cruz, Christopher
Publié: (2025)

EVM-QuestBench: An Execution-Grounded Benchmark for Natural-Language Transaction Code Generation
par: Yang, Pei, et autres
Publié: (2026)