arXiv subjects

Yutong Han

Publications and source records attributed to Yutong Han.

3 recordsLinked to original sources

arXiv2026

Interactive Benchmarks

Existing reasoning evaluation paradigms suffer from different limitations: fixed benchmarks are increasingly saturated and vulnerable to contamination, while preference-based evaluations rely on subjective judgments. We argue that a core aspect of intelligence is the ability to decide what information to acquire and how to use it effectively. We propose Interactive Benchmarks, a unified evaluation paradigm that assesses a model's reasoning ability through budgeted multi-turn interaction. We evaluate models under this framework in two settings: Interactive Proofs, where models interact with a judge to solve Logic, UI2Html, and Mathematics tasks under objective feedback; and Interactive Games, where models reason strategically to maximize long-horizon utilities. Our results show that interactive benchmarks provide a more robust assessment of this dimension of model intelligence, revealing substantial room for improvement in interactive scenarios.

cs.AI↗

arXiv2026

On Diagonalizable Systems with Random Structure

Diagonalizability plays an important role in the analysis and design of multivariable systems. A structured matrix is called structurally diagonalizable if almost all of its numerical realizations, obtained by assigning real values to its free entries, are diagonalizable. Structural diagonalizability is useful for the verification and optimization of various structural system properties. In this paper, we study the asymptotic probability distribution of structural diagonalizability for structured systems whose system matrices are represented by directed Erd\H{o}s-R\'{e}nyi random graphs. Leveraging a recently established graph-theoretic characterization of structural diagonalizability, we analyze the distribution of structurally diagonalizable graphs under different edge-density regimes. For dense graphs, we prove that the system is almost always structurally diagonalizable. For graphs of medium density, we derive tight upper and lower bounds on the asymptotic probability of structural diagonalizability. For extremely sparse graphs, we show that this probability approaches 0. The theoretical results are validated through extensive numerical simulations with varying numbers of vertices and connection probabilities.

math.OC↗

arXiv2024

A Dual-View Approach to Classifying Radiology Reports by Co-Training

Radiology report analysis provides valuable information that can aid with public health initiatives, and has been attracting increasing attention from the research community. In this work, we present a novel insight that the structure of a radiology report (namely, the Findings and Impression sections) offers different views of a radiology scan. Based on this intuition, we further propose a co-training approach, where two machine learning models are built upon the Findings and Impression sections, respectively, and use each other's information to boost performance with massive unlabeled data in a semi-supervised manner. We conducted experiments in a public health surveillance study, and results show that our co-training approach is able to improve performance using the dual views and surpass competing supervised and semi-supervised methods.

cs.CL↗