arXiv subjects

Dongseok Lee

Publications and source records attributed to Dongseok Lee.

10 recordsLinked to original sources

arXiv2026

Training-free LLM Verification via Recycling Few-shot Examples

Although large language models (LLMs) have achieved remarkable performance, the inherent stochasticity of their reasoning processes and varying conclusions present significant challenges. Majority voting or Best-of-N with external verifiers has been explored to mitigate this, but these approaches are limited in applicability or require additional training. To address this problem, we propose a novel framework that Recycles Few-shot examples to verify LLM outputs (ReFeri). Our key idea is to utilize the given few-shot examples not only to generate outputs, but also to evaluate the candidate outputs. Specifically, ReFeri combines a forward confidence score with a backward reconstruction penalty to select candidates that follow few-shot guidance while avoiding demonstration-specific overfitting. Experiments with three different LLMs across seven diverse tasks demonstrate that our framework significantly improves the accuracy of LLMs---achieving an average relative gain of 8.2%---through effective response selection.

cs.LG↗

arXiv2026

LaRA: Layer-wise Representation Analysis for Detecting Data Contamination in RL Post-Training

Reinforcement learning (RL) post-training has shown to improve reasoning in large language models (LLMs). However, there has been little exploration on the problem of data contamination in RL post-training, potentially undermining generalization and evaluation reliability of the training process itself. Existing detection methods primarily rely on output-level signals such as likelihood or entropy, which become unreliable for RL-trained models since RL shapes behavior through trajectory-level rewards rather than token likelihoods. We propose LaRA, a layer-wise representation analysis framework for detecting contamination in RL post-trained LLMs. LaRA introduces three complementary metrics, measuring perturbation sensitivity, directional collapse, and local representation rigidity under controlled perturbations. We find that contamination produces progressive geometric deviations across layers, including amplified perturbation sensitivity, stronger directional collapse, and enhanced local rigidity. Based on our findings, we also develop a contamination detection protocol that aggregates representation-level deviations across layers and metrics. Experiments on RL-trained reasoning models show that our protocol outperforms existing output-level baselines for contamination detection.

cs.LG↗

arXiv2026

LookBack: Where and How to Score LVLM Responses via Visual Reference Usage

Large Vision-Language Models (LVLMs) integrate visual perception with language generation, enabling responses that span image understanding and complex reasoning. However, LVLMs do not just inherit the text-level hallucinations; they also hallucinate against the image, producing fluent responses ungrounded in what they see. This makes LVLM response scoring inherently harder, and our diagnostics show that existing confidence-based metrics adopted from LLMs are insufficient for LVLMs. Specifically, removing the input image barely changes confidence-based selection, suggesting that output-space confidence primarily captures textual plausibility rather than agreement with the image. To address this gap, we propose LookBack, a training-free LVLM response scoring method that augments token likelihood with visual lookback score, a lightweight measure of how strongly each response token refers to image tokens. Across four benchmarks and three models, LookBack consistently improves Best-of-$N$ selection over existing baselines with negligible additional overhead.

cs.CV↗

arXiv2026

The Amazing Agent Race: Strong Tool Users, Weak Navigators

Existing tool-use benchmarks for LLM agents are overwhelmingly linear: our analysis of six benchmarks shows 55 to 100% of instances are simple chains of 2 to 5 steps. We introduce The Amazing Agent Race (AAR), a benchmark featuring directed acyclic graph (DAG) puzzles (or "legs") with fork-merge tool chains. We release 1,400 instances across two variants: sequential (800 legs) and compositional (600 DAG legs). Agents must navigate Wikipedia, execute multi-step tool chains, and aggregate results into a verifiable answer. Legs are procedurally generated from Wikipedia seeds across four difficulty levels with live-API validation. Three complementary metrics (finish-line accuracy, pit-stop visit rate, and roadblock completion rate) separately diagnose navigation, tool-use, and arithmetic failures. Evaluating three agent frameworks on 1,400 legs, the best achieves only 37.2% accuracy. Navigation errors dominate (27 to 52% of trials) while tool-use errors remain below 17%, and agent architecture matters as much as model scale (Claude Code matches Codex CLI at 37% with 6x fewer tokens). The compositional structure of AAR reveals that agents fail not at calling tools but at navigating to the right pages, a blind spot invisible to linear benchmarks. The project page can be accessed at: https://minnesotanlp.github.io/the-amazing-agent-race

cs.AI↗

arXiv2025

Latent-Space Mean-Field Theory for Deep BitNet-like Training: Constrained Gradient Flows with Smooth Quantization and STE Limits

This work develops a mean-field analysis for the asymptotic behavior of deep BitNet-like architectures as smooth quantization parameters approach zero. We establish that empirical measures of latent weights converge weakly to solutions of constrained continuity equations under vanishing quantization smoothing. Our main theoretical contribution demonstrates that the natural exponential decay in smooth quantization cancels out apparent singularities, yielding uniform bounds on mean-field dynamics independent of smoothing parameters. Under standard regularity assumptions, we prove convergence to a well-defined limit that provides the mathematical foundation for gradient-based training of quantized neural networks through distributional analysis.

math.OC↗

arXiv2025

Toward a Robust and Generalizable Metamaterial Foundation Model

Advances in material functionalities drive innovations across various fields, where metamaterials-defined by structure rather than composition-are leading the way. Despite the rise of artificial intelligence (AI)-driven design strategies, their impact is limited by task-specific retraining, poor out-of-distribution(OOD) generalization, and the need for separate models for forward and inverse design. To address these limitations, we introduce the Metamaterial Foundation Model (MetaFO), a Bayesian transformer-based foundation model inspired by large language models. MetaFO learns the underlying mechanics of metamaterials, enabling probabilistic, zero-shot predictions across diverse, unseen combinations of material properties and structural responses. It also excels in nonlinear inverse design, even under OOD conditions. By treating metamaterials as an operator that maps material properties to structural responses, MetaFO uncovers intricate structure-property relationships and significantly expands the design space. This scalable and generalizable framework marks a paradigm shift in AI-driven metamaterial discovery, paving the way for next-generation innovations.

cs.CE↗

arXiv2024

MaD-Scientist: AI-based Scientist solving Convection-Diffusion-Reaction Equations Using Massive PINN-Based Prior Data

Large language models (LLMs), like ChatGPT, have shown that even trained with noisy prior data, they can generalize effectively to new tasks through in-context learning (ICL) and pre-training techniques. Motivated by this, we explore whether a similar approach can be applied to scientific foundation models (SFMs). Our methodology is structured as follows: (i) we collect low-cost physics-informed neural network (PINN)-based approximated prior data in the form of solutions to partial differential equations (PDEs) constructed through an arbitrary linear combination of mathematical dictionaries; (ii) we utilize Transformer architectures with self and cross-attention mechanisms to predict PDE solutions without knowledge of the governing equations in a zero-shot setting; (iii) we provide experimental evidence on the one-dimensional convection-diffusion-reaction equation, which demonstrate that pre-training remains robust even with approximated prior data, with only marginal impacts on test accuracy. Notably, this finding opens the path to pre-training SFMs with realistic, low-cost data instead of (or in conjunction with) numerical high-cost data. These results support the conjecture that SFMs can improve in a manner similar to LLMs, where fully cleaning the vast set of sentences crawled from the Internet is nearly impossible.

cs.LG↗

arXiv2024

Singular Layer Physics-Informed Neural Network Method for Convection-Dominated Boundary Layer Problems in Two Dimensions

This research explores neural network-based numerical approximation of two-dimensional convection-dominated singularly perturbed problems on square, circular, and elliptic domains. Singularly perturbed boundary value problems pose significant challenges due to sharp boundary layers in their solutions. Additionally, the characteristic points of these domains give rise to degenerate boundary layer problems. The stiffness of these problems, caused by sharp singular layers, can lead to substantial computational errors if not properly addressed. Conventional neural network-based approaches often fail to capture these sharp transitions accurately, highlighting a critical flaw in machine learning methods. To address these issues, we conduct a thorough boundary layer analysis to enhance our understanding of sharp transitions within the boundary layers, guiding the application of numerical methods. Specifically, we employ physics-informed neural networks (PINNs) to better handle these boundary layer problems. However, PINNs may struggle with rapidly varying singularly perturbed solutions in small domain regions, leading to inaccurate or unstable results. To overcome this limitation, we introduce a semi-analytic method that augments PINNs with singular layers or corrector functions. Our numerical experiments demonstrate significant improvements in both accuracy and stability, showcasing the effectiveness of our proposed approach.

math.NA↗

arXiv2024

Simple arithmetic operation in latent space can generate a novel three dimensional graph metamaterials

Recent advancements in artificial intelligence (AI)-based design strategies for metamaterials have revolutionized the creation of customizable architectures spanning nano- to macro-scale dimensions, achieving unprecedented mechanical behaviors that surpass the inherent properties of the constituent materials. However, the growing complexity of these methods poses challenges in generating diverse metamaterials without substantial human and computational resources, hindering widespread adoption. Addressing this, our study introduces an innovative design strategy capable of generating various three-dimensional graph metamaterials using simple arithmetic operations within the latent space. By seamlessly integrating hidden representations of disentangled latent space and latent diffusion processes, our approach provides a comprehensive understanding of complex design spaces, generating diverse graph metamaterials through arithmetic operations. This methodology stands as a versatile tool for creating structures ranging from repetitive lattice structures to functionally graded mechanical metamaterials. It also serves as an inverse design strategy for diverse lattice structures, including crystalline structures and those made of trabecular bone. We believe that this methodology represents a foundational step in advancing our comprehension of the intricate latent design space, offering the potential to establish a unified model for various traditional generative models in the realm of mechanical metamaterials.

physics.app-ph↗

arXiv2019

Cooperative Multi-Agent Reinforcement Learning Framework for Scalping Trading

We explore deep Reinforcement Learning(RL) algorithms for scalping trading and knew that there is no appropriate trading gym and agent examples. Thus we propose gym and agent like Open AI gym in finance. Not only that, we introduce new RL framework based on our hybrid algorithm which leverages between supervised learning and RL algorithm and uses meaningful observations such order book and settlement data from experience watching scalpers trading. That is very crucial information for traders behavior to be decided. To feed these data into our model, we use spatio-temporal convolution layer, called Conv3D for order book data and temporal CNN, called Conv1D for settlement data. Those are preprocessed by episode filter we developed. Agent consists of four sub agents divided to clarify their own goal to make best decision. Also, we adopted value and policy based algorithm to our framework. With these features, we could make agent mimic scalpers as much as possible. In many fields, RL algorithm has already begun to transcend human capabilities in many domains. This approach could be a starting point to beat human in the financial stock market, too and be a good reference for anyone who wants to design RL algorithm in real world domain. Finally, weexperiment our framework and gave you experiment progress.

cs.AI↗