1 citations · 1 across the 18 of their papers we have counts for
22 papers
Who Is the Influencer? Consensus Dynamics in Multi-Agent Systems
Gouki Minegishi, Jingyuan Feng, Hiroki Furuta +3
Multi-agent LLM systems are increasingly used in practice, but their behavior is complex and hard to understand. We design a simple task to ask which model has the most influence o…
Matching Object or Relation? Tracing Abstract Reasoning Inside VLMs
Gouki Minegishi, Hiroki Furuta, Takeshi Kojima +2
Vision Language Models (VLMs) excel on visual benchmarks but fail systematically on tasks requiring abstract reasoning. Existing benchmarks document this failure but cannot say \em…
Anosognosia in LLMs: Probing Self-Awareness of Quantized Computational Substrate
Yoshihiro Izawa, Gouki Minegishi, Yoko Yamakata
Can LLMs recognize degradation in their own computational substrate? Inspired by anosognosia, a neurological condition in which patients fail to recognize impairments in their own…
Few-Shot Demonstrations Elicit the Use of In-Context World Representations in LLMs
Kohsei Matsutani, Gouki Minegishi, Core Francisco Park +3
Large language models (LLMs), when acting as agents, are expected to take observed data in context, infer the latent state space underlying the world, and leverage it for downstrea…
In-Context Neurofeedback: Can LLMs Control Their Internal Representations through Privileged Access?
Koshiro Aoki, Ryota Takatsuki, Gouki Minegishi +2
Whether large language models (LLMs) can control their own internal representations matters for both machine metacognition and AI safety. A recent study applied neurofeedback to LL…
What Keeps Vision-Language Models Looking at the Image?
Hiroto Osaka, Shohei Taniguchi, Gouki Minegishi +3
When do vision-language models need direct access to the image while generating an answer? We study image dependence during answer generation by examining how the visual informatio…
On Advantage Estimates for Max@K Policy Gradients
Shota Takashiro, Soichiro Nishimori, Paavo Parmas +6
Reinforcement learning with verifiable rewards is widely used for post-training reasoning models, but sparse outcome rewards make exploration difficult. A complementary approach is…
Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
Kohsei Matsutani, Gouki Minegishi, Takeshi Kojima +2
Large language models (LLMs) can now solve complex problems through long chain-of-thought (CoT) reasoning, but the trade-off between performance and token cost remains a central ch…
LIT-RAGBench: Benchmarking Generator Capabilities of Large Language Models in Retrieval-Augmented Generation
Koki Itai, Shunichi Hasegawa, Yuta Yamamoto +2
Retrieval-Augmented Generation (RAG) is a framework in which a Generator, such as a Large Language Model (LLM), produces answers by retrieving documents from an external collection…
Understanding Emergent Misalignment via Feature Superposition Geometry
Gouki Minegishi, Hiroki Furuta, Takeshi Kojima +2
Emergent misalignment, where fine-tuning on narrow, non-harmful tasks induces harmful behaviors, poses a key challenge for AI safety in LLMs. Despite growing empirical evidence, it…
Safe Transformer: An Explicit Safety Bit For Interpretable And Controllable Alignment
Jingyuan Feng, Andrew Gambardella, Gouki Minegishi +3
Current safety alignment methods encode safe behavior implicitly within model parameters, creating a fundamental opacity: we cannot easily inspect why a model refuses a request, no…
Steering at the Source: Style Modulation Heads for Robust Persona Control
Yoshihiro Izawa, Gouki Minegishi, Koshi Eguchi +2
Activation steering offers a computationally efficient mechanism for controlling Large Language Models (LLMs) without fine-tuning. While effectively controlling target traits (e.g.…
Emergent Analogical Reasoning in Transformers
Gouki Minegishi, Jingyuan Feng, Hiroki Furuta +3
Analogy is a central faculty of human intelligence, enabling abstract patterns discovered in one domain to be applied to another. Despite its central role in cognition, the mechani…
Interpreting Multi-Attribute Confounding through Numerical Attributes in Large Language Models
Hirohane Takagi, Gouki Minegishi, Shota Kizawa +2
Although behavioral studies have documented numerical reasoning errors in large language models (LLMs), the underlying representational mechanisms remain unclear. We hypothesize th…
RL Squeezes, SFT Expands: A Comparative Study of Reasoning LLMs
Kohsei Matsutani, Shota Takashiro, Gouki Minegishi +3
Large language models (LLMs) are typically trained by reinforcement learning (RL) with verifiable rewards (RLVR) and supervised fine-tuning (SFT) on reasoning traces to improve the…
Mechanism of Task-oriented Information Removal in In-context Learning
Hakaze Cho, Haolin Yang, Gouki Minegishi +1
In-context Learning (ICL) is an emerging few-shot learning paradigm based on modern Language Models (LMs), yet its inner mechanism remains unclear. In this paper, we investigate th…
Beyond Induction Heads: In-Context Meta Learning Induces Multi-Phase Circuit Emergence
Gouki Minegishi, Hiroki Furuta, Shohei Taniguchi +2
Transformer-based language models exhibit In-Context Learning (ICL), where predictions are made adaptively based on context. While prior work links induction heads to ICL through a…
Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties
Gouki Minegishi, Hiroki Furuta, Takeshi Kojima +2
Recent large-scale reasoning models have achieved state-of-the-art performance on challenging mathematical benchmarks, yet the internal mechanisms underlying their success remain p…
Rethinking Evaluation of Sparse Autoencoders through the Representation of Polysemous Words
Gouki Minegishi, Hiroki Furuta, Yusuke Iwasawa +1
Sparse autoencoders (SAEs) have gained a lot of attention as a promising tool to improve the interpretability of large language models (LLMs) by mapping the complex superposition o…
ADOPT: Modified Adam Can Converge with Any with the Optimal Rate
Shohei Taniguchi, Keno Harada, Gouki Minegishi +7
Adam is one of the most popular optimization algorithms in deep learning. However, it is known that Adam does not converge in theory unless choosing a hyperparameter, i.e., ,…
Towards Empirical Interpretation of Internal Circuits and Properties in Grokked Transformers on Modular Polynomials
Hiroki Furuta, Gouki Minegishi, Yusuke Iwasawa +1
Grokking has been actively explored to reveal the mystery of delayed generalization and identifying interpretable representations and algorithms inside the grokked models is a sugg…
Bridging Lottery Ticket and Grokking: Understanding Grokking from Inner Structure of Networks
Gouki Minegishi, Yusuke Iwasawa, Yutaka Matsuo
Grokking is an intriguing phenomenon of delayed generalization, where neural networks initially memorize training data with perfect accuracy but exhibit poor generalization, subseq…