Picture for Kentaro Inui

Kentaro Inui

MBZUAI, Tohoku University, RIKEN

How LLMs Learn: Tracing Internal Representations with Sparse Autoencoders

Add code
Mar 09, 2025
Viaarxiv icon

Syntactic Learnability of Echo State Neural Language Models at Scale

Add code
Mar 03, 2025
Viaarxiv icon

Rectifying Belief Space via Unlearning to Harness LLMs' Reasoning

Add code
Feb 28, 2025
Viaarxiv icon

Number Representations in LLMs: A Computational Parallel to Human Perception

Add code
Feb 22, 2025
Viaarxiv icon

Large Language Models Are Human-Like Internally

Add code
Feb 03, 2025
Figure 1 for Large Language Models Are Human-Like Internally
Figure 2 for Large Language Models Are Human-Like Internally
Figure 3 for Large Language Models Are Human-Like Internally
Figure 4 for Large Language Models Are Human-Like Internally
Viaarxiv icon

Automatic Feedback Generation for Short Answer Questions using Answer Diagnostic Graphs

Add code
Jan 27, 2025
Viaarxiv icon

Weight-based Analysis of Detokenization in Language Models: Understanding the First Stage of Inference Without Inference

Add code
Jan 27, 2025
Figure 1 for Weight-based Analysis of Detokenization in Language Models: Understanding the First Stage of Inference Without Inference
Figure 2 for Weight-based Analysis of Detokenization in Language Models: Understanding the First Stage of Inference Without Inference
Figure 3 for Weight-based Analysis of Detokenization in Language Models: Understanding the First Stage of Inference Without Inference
Figure 4 for Weight-based Analysis of Detokenization in Language Models: Understanding the First Stage of Inference Without Inference
Viaarxiv icon

Think-to-Talk or Talk-to-Think? When LLMs Come Up with an Answer in Multi-Step Reasoning

Add code
Dec 02, 2024
Viaarxiv icon

Repetition Neurons: How Do Language Models Produce Repetitions?

Add code
Oct 17, 2024
Viaarxiv icon

The Geometry of Numerical Reasoning: Language Models Compare Numeric Properties in Linear Subspaces

Add code
Oct 17, 2024
Viaarxiv icon