arXiv subjects

Junwon Ko

Publications and source records attributed to Junwon Ko.

4 recordsLinked to original sources

arXiv2026

Direct Diversity Optimization for Diverse Successful Trajectories in Preference Post-Training

LLM agents for sequential decision tasks are often post-trained with trajectory-level outcome labels, but such labels provide little supervision for preserving multiple successful branches from the same decision state. We study this problem as successful strategy coverage: how broadly a model realizes distinct successful strategies under a fixed rollout budget. We present Direct Diversity Optimization (DDO), an offline post-training method that combines Divergence-Tree Collection (DTC) with the Reference-Relative Target-Odds Objective (RTO). DTC constructs state-aligned branch sets rooted at shared decision states, and RTO trains the model to match reference-relative targets over successful alternatives. DDO achieves the strongest task success and successful strategy coverage among the compared post-training methods across BabyAI, BabaIsAI, and WebShop. It also achieves the highest recovery rate after local action replacement and higher task success and coverage than successful-only imitation and decoding-time diversification controls.

cs.CL↗

arXiv2025

Preference Distillation via Value based Reinforcement Learning

Direct Preference Optimization (DPO) is a powerful paradigm to align language models with human preferences using pairwise comparisons. However, its binary win-or-loss supervision often proves insufficient for training small models with limited capacity. Prior works attempt to distill information from large teacher models using behavior cloning or KL divergence. These methods often focus on mimicking current behavior and overlook distilling reward modeling. To address this issue, we propose \textit{Teacher Value-based Knowledge Distillation} (TVKD), which introduces an auxiliary reward from the value function of the teacher model to provide a soft guide. This auxiliary reward is formulated to satisfy potential-based reward shaping, ensuring that the global reward structure and optimal policy of DPO are preserved. TVKD can be integrated into the standard DPO training framework and does not require additional rollouts. Our experimental results show that TVKD consistently improves performance across various benchmarks and model sizes.

cs.CL↗

arXiv2025

SFLD: Reducing the content bias for AI-generated Image Detection

Identifying AI-generated content is critical for the safe and ethical use of generative AI. Recent research has focused on developing detectors that generalize to unknown generators, with popular methods relying either on high-level features or low-level fingerprints. However, these methods have clear limitations: biased towards unseen content, or vulnerable to common image degradations, such as JPEG compression. To address these issues, we propose a novel approach, SFLD, which incorporates PatchShuffle to integrate high-level semantic and low-level textural information. SFLD applies PatchShuffle at multiple levels, improving robustness and generalization across various generative models. Additionally, current benchmarks face challenges such as low image quality, insufficient content preservation, and limited class diversity. In response, we introduce TwinSynths, a new benchmark generation methodology that constructs visually near-identical pairs of real and synthetic images to ensure high quality and content preservation. Our extensive experiments and analysis show that SFLD outperforms existing methods on detecting a wide variety of fake images sourced from GANs, diffusion models, and TwinSynths, demonstrating the state-of-the-art performance and generalization capabilities to novel generative models.

cs.CV↗

arXiv2024

AH-OCDA: Amplitude-based Curriculum Learning and Hopfield Segmentation Model for Open Compound Domain Adaptation

Open compound domain adaptation (OCDA) is a practical domain adaptation problem that consists of a source domain, target compound domain, and unseen open domain. In this problem, the absence of domain labels and pixel-level segmentation labels for both compound and open domains poses challenges to the direct application of existing domain adaptation and generalization methods. To address this issue, we propose Amplitude-based curriculum learning and a Hopfield segmentation model for Open Compound Domain Adaptation (AH-OCDA). Our method comprises two complementary components: 1) amplitude-based curriculum learning and 2) Hopfield segmentation model. Without prior knowledge of target domains within the compound domains, amplitude-based curriculum learning gradually induces the semantic segmentation model to adapt from the near-source compound domain to the far-source compound domain by ranking unlabeled compound domain images through Fast Fourier Transform (FFT). Additionally, the Hopfield segmentation model maps segmentation feature distributions from arbitrary domains to the feature distributions of the source domain. AH-OCDA achieves state-of-the-art performance on two OCDA benchmarks and extended open domains, demonstrating its adaptability to continuously changing compound domains and unseen open domains.

cs.AI↗