arXiv subjects

Yusuke Takase

Publications and source records attributed to Yusuke Takase.

9 recordsLinked to original sources

arXiv2026

Language Model Maps for Prompt-Response Distributions via Log-Likelihood Vectors

We propose a method that represents language models by log-likelihood vectors over prompt-response pairs and constructs model maps for comparing their conditional distributions. In this space, squared Euclidean distances between models are approximately proportional to the KL divergence between the corresponding conditional distributions. Experiments on a large collection of publicly available language models show that the maps capture meaningful global structure, including relationships to model attributes and task performance. The representation also captures systematic shifts induced by prompt modifications and their approximate additive compositionality; we use the vectors to predict downstream task scores and leverage their additive structure to approximate the effects of composite prompt operations without directly observing the corresponding log-likelihood vectors. We further introduce PMI vectors to reduce the influence of unconditional distributions; in some cases, PMI-based model maps better reflect training-data-related differences. Overall, the framework supports the analysis and prediction of input-dependent model behavior.

cs.CL↗

arXiv2026

Measuring Affinity between Attention-Head Weight Subspaces via the Projection Kernel

Understanding relationships between attention heads is essential for interpreting the internal structure of Transformers, yet existing metrics do not capture this structure well. We focus on the subspaces spanned by attention-head weight matrices and quantify head-to-head relationships using the Projection Kernel (PK), a principal-angle-based measure of subspace similarity. Experiments show that PK reproduces known head-to-head interactions on the IOI task more clearly than prior metrics such as the Composition Score. We further introduce a framework to quantify the informativeness of PK distributions by comparing them with a reference distribution derived from random orthogonal subspaces. As an application, we analyze a directed graph constructed from PK and show that, in GPT2-small, L4H7 acts as a hub by functioning as an identity head.

cs.CL↗

arXiv2025

Establishing a Scale for Kullback-Leibler Divergence in Language Models Across Various Settings

Log-likelihood vectors define a common space for comparing language models as probability distributions, enabling unified comparisons across heterogeneous settings. We extend this framework to training checkpoints and intermediate layers, and establish a consistent scale for KL divergence across pretraining, model size, random seeds, quantization, fine-tuning, and layers. Analysis of Pythia pretraining trajectories further shows that changes in log-likelihood space, as measured by the scaling behavior of KL divergence, are much smaller than in weight space, resulting in subdiffusive learning trajectories and early stabilization of language-model behavior despite weight drift.

cs.CL↗

arXiv2025

Design of the full-sky scanning strategy and systematic effect control in a cosmic microwave background probe

The quest for primordial $B$-mode polarization signatures in the Cosmic Microwave Background (CMB) is a major goal of contemporary cosmology. Detecting these signatures would confirm primordial gravitational waves and allow precise determination of the tensor-to-scalar ratio, $r$, which is crucial for distinguishing between inflationary models. This requires high-precision, full-sky observations from space-based platforms to avoid atmospheric fluctuations and windows. Since $B$-mode signatures are much fainter than CMB temperature anisotropies, their detection requires well-designed calibration and mitigation strategies. Next-generation space observatories such as LiteBIRD, which use Half-Wave Plate (HWP) modulation technology, represent a significant advance. This technology allows single-detector observations, eliminating the need for pair-differential detection and its associated systematic complexities. The study begins with the optimization of scanning strategy parameters for missions with HWP to improve in-flight calibration, suppress systematic effects, and develop robust null-test methods. In addition, we present a method for calculating the impact of systematic effects on the estimation of $r$ using the SBM (Spin-Based Map-making) software framework, which performs fast map-making in spin space. This method allows the decomposition and elimination of systematic effects in terms of spins, and we demonstrate the elimination of several systematic effects.

astro-ph.CO↗

arXiv2025

DeLTa: A Decoding Strategy based on Logit Trajectory Prediction Improves Factuality and Reasoning Ability

Large Language Models (LLMs) are increasingly being used in real-world applications. However, concerns about the reliability of the content they generate persist, as it frequently deviates from factual correctness or exhibits deficiencies in logical reasoning. This paper proposes a novel decoding strategy aimed at enhancing both factual accuracy and inferential reasoning without requiring any modifications to the architecture or pre-trained parameters of LLMs. Our approach adjusts next-token probabilities by analyzing the trajectory of logits from lower to higher layers in Transformers and applying linear regression. We find that this Decoding by Logit Trajectory-based approach (DeLTa) effectively reinforces factuality and reasoning while mitigating incorrect generation. Experiments on TruthfulQA demonstrate that DeLTa attains up to a 4.9% improvement over the baseline. Furthermore, it enhances performance by up to 8.1% on StrategyQA and 7.3% on GSM8K, both of which demand strong reasoning capabilities.

cs.CL↗

arXiv2025

Mapping 1,000+ Language Models via the Log-Likelihood Vector

To compare autoregressive language models at scale, we propose using log-likelihood vectors computed on a predefined text set as model features. This approach has a solid theoretical basis: when treated as model coordinates, their squared Euclidean distance approximates the Kullback-Leibler divergence of text-generation probabilities. Our method is highly scalable, with computational cost growing linearly in both the number of models and text samples, and is easy to implement as the required features are derived from cross-entropy loss. Applying this method to over 1,000 language models, we constructed a "model map," providing a new perspective on large-scale model analysis.

cs.CL↗

arXiv2024

Axis Tour: Word Tour Determines the Order of Axes in ICA-transformed Embeddings

Word embedding is one of the most important components in natural language processing, but interpreting high-dimensional embeddings remains a challenging problem. To address this problem, Independent Component Analysis (ICA) is identified as an effective solution. ICA-transformed word embeddings reveal interpretable semantic axes; however, the order of these axes are arbitrary. In this study, we focus on this property and propose a novel method, Axis Tour, which optimizes the order of the axes. Inspired by Word Tour, a one-dimensional word embedding method, we aim to improve the clarity of the word embedding space by maximizing the semantic continuity of the axes. Furthermore, we show through experiments on downstream tasks that Axis Tour yields better or comparable low-dimensional embeddings compared to both PCA and ICA.

cs.CL↗

arXiv2023

Continuous optimization by quantum adaptive distribution search

In this paper, we introduce the quantum adaptive distribution search (QuADS), a quantum continuous optimization algorithm that integrates Grover adaptive search (GAS) with the covariance matrix adaptation - evolution strategy (CMA-ES), a classical technique for continuous optimization. QuADS utilizes the quantum-based search capabilities of GAS and enhances them with the principles of CMA-ES for more efficient optimization. It employs a multivariate normal distribution for the initial state of the quantum search and repeatedly updates it throughout the optimization process. Our numerical experiments show that QuADS outperforms both GAS and CMA-ES. This is achieved through adaptive refinement of the initial state distribution rather than consistently using a uniform state, resulting in fewer oracle calls. This study presents an important step toward exploiting the potential of quantum computing for continuous optimization.

quant-ph↗

arXiv2023

Zero-Shot Edge Detection with SCESAME: Spectral Clustering-based Ensemble for Segment Anything Model Estimation

This paper proposes a novel zero-shot edge detection with SCESAME, which stands for Spectral Clustering-based Ensemble for Segment Anything Model Estimation, based on the recently proposed Segment Anything Model (SAM). SAM is a foundation model for segmentation tasks, and one of the interesting applications of SAM is Automatic Mask Generation (AMG), which generates zero-shot segmentation masks of an entire image. AMG can be applied to edge detection, but suffers from the problem of overdetecting edges. Edge detection with SCESAME overcomes this problem by three steps: (1) eliminating small generated masks, (2) combining masks by spectral clustering, taking into account mask positions and overlaps, and (3) removing artifacts after edge detection. We performed edge detection experiments on two datasets, BSDS500 and NYUDv2. Although our zero-shot approach is simple, the experimental results on BSDS500 showed almost identical performance to human performance and CNN-based methods from seven years ago. In the NYUDv2 experiments, it performed almost as well as recent CNN-based methods. These results indicate that our method effectively enhances the utility of SAM and can be a new direction in zero-shot edge detection methods.

cs.CV↗