Gouki Minegishi
Scholar

Gouki Minegishi

Google Scholar ID: sxGpoYMAAAAJ
University of Tokyo
Deep LearningInterpretability
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
45
 
H-index
4
 
i10-index
2
 
Publications
13
 
Co-authors
0
 
Contact
Emailminegishi@weblab.t.u-tokyo.ac.jpCVOpen ↗TwitterOpen ↗GitHubOpen ↗
Publications
19 items
Matching Object or Relation? Tracing Abstract Reasoning Inside VLMs
2026
Cited
0
Anosognosia in LLMs: Probing Self-Awareness of Quantized Computational Substrate
2026
Cited
0
Few-Shot Demonstrations Elicit the Use of In-Context World Representations in LLMs
2026
Cited
0
In-Context Neurofeedback: Can LLMs Control Their Internal Representations through Privileged Access?
2026
Cited
0
Visual Access Boundaries in Vision-Language Model Reasoning
2026
Cited
0
On Advantage Estimates for Max@K Policy Gradients
2026
Cited
0
Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
2026
Cited
0
Safe Transformer: An Explicit Safety Bit For Interpretable And Controllable Alignment
2026
Cited
0
Resume
Academic Achievements
  • - Neurips2025: Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties
  • - ICML2025: Beyond Induction Heads: In-Context Meta Learning Induces Multi-Phase Circuit Emergence
  • - ICLR2025: Rethinking Evaluation of Sparse Autoencoders through the Representation of Polysemous Words
Education
  • First-year PhD student at The University of Tokyo, mentored by Professor Yutaka Matsuo.
Background
  • Research interest: mechanistic interpretability, aiming to unravel the internal mechanisms that drive today's AI systems, with the ultimate goal of understanding what truly constitutes human intelligence.