0

Yonatan Belinkov

Papers
21

Cite

Notes

Only stored in your browser.

Attribution

Affiliations & profile
Semantic Scholar
Attribution policy →
21papers

Authored papers

21

SAEs Are Good for Steering -- If You Select the Right Features

arXiv 2025

2025

Follow the Flow: On Information Flow Across Textual Tokens in Text-to-Image Models

arXiv 2025

2025

Inside-Out: Hidden Factual Knowledge in LLMs

arXiv 2025

2025

Position-aware Automatic Circuit Discovery

arXiv 2025

2025

Have Faith in Faithfulness: Going Beyond Circuit Overlap When Finding Model Mechanisms

arXiv 2024

2024

Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models

arXiv 2024

2024

LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations

arXiv 2024

2024

Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics

arXiv 2024

2024

Distinguishing Ignorance from Error in LLM Hallucinations

arXiv 2024

2024

Confidence Regulation Neurons in Language Models

arXiv 2024

2024

Jamba-1.5: Hybrid Transformer-Mamba Models at Scale

arXiv 2024

2024

Semantics and Spatiality of Emergent Communication

arXiv 2024

2024

Generating Benchmarks for Factuality Evaluation of Language Models

arXiv 2023

2023

VISIT: Visualizing and Interpreting the Semantic Information Flow of Transformers

arXiv 2023

2023

A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis

arXiv 2023

2023

Instructed to Bias: Instruction-Tuned Language Models Exhibit Emergent Cognitive Bias

arXiv 2023

2023

Unified Concept Editing in Diffusion Models

arXiv 2023

2023

Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models

TMLR

2022

Locating and Editing Factual Associations in GPT

arXiv 2022

2022

Mass-Editing Memory in a Transformer

arXiv 2022

2022

Improving Neural Language Models by Segmenting, Attending, and Predicting the Future

improving-neural-language-models-by-1

2019

Affiliations

No known affiliations.

Frequent co-authors

10

from 21 papers