Mirae Kim

Eunbyeol Cho, Yunseung Lee, Mirae Kim et al.

Large Language Models (LLMs) have advanced financial automation through Retrieval-Augmented Generation (RAG), yet hallucinations remain a critical barrier to deployment in high-stakes environments. Existing benchmarks focus on single-turn, English-centric tasks, leaving the multi-turn dynamics and linguistic-regulatory nuances of the Korean financial domain unaddressed. We introduce K-FinHallu, the first benchmark for hallucination detection in multi-turn Korean financial RAG. We construct multi-turn dialogues from authentic Korean financial documents and inject hallucinations under a proposed hierarchical taxonomy based on context answerability that explicitly accounts for justified abstention. Benchmarking frontier and open-source LLMs as hallucination detectors, we find that even the strongest models struggle with fine-grained financial diagnostics and refusal behavior. While fine-tuning an 8B model on our training split yields performance competitive with frontier LLMs, justified abstention remains the weakest axis across all evaluated models.

Mirae Kim, Simon Woo

Information has exploded on the Internet and mobile with the advent of the big data era. In particular, recommendation systems are widely used to help consumers who struggle to select the best products among such a large amount of information. However, recommendation systems are vulnerable to malicious user biases, such as fake reviews to promote or demote specific products, as well as attacks that steal personal information. Such biases and attacks compromise the fairness of the recommendation model and infringe the privacy of users and systems by distorting data.Recently, deep-learning collaborative filtering recommendation systems have shown to be more vulnerable to this bias. In this position paper, we examine the effects of bias that cause various ethical and social issues, and discuss the need for designing the robust recommendation system for fairness and stability.

Mirae Kim, Seonghun Jeong, Youngjun Kwak

Jailbreaking poses a significant risk to the deployment of Large Language Models (LLMs) and Vision Language Models (VLMs). VLMs are particularly vulnerable because they process both text and images, creating broader attack surfaces. However, available resources for jailbreak detection are scarce, particularly in finance. To address this gap, we present FENCE, a bilingual (Korean-English) multimodal dataset for training and evaluating jailbreak detectors in financial applications. FENCE emphasizes domain realism through finance-relevant queries paired with image-grounded threats. Experiments with commercial and open-source VLMs reveal consistent vulnerabilities, with GPT-4o showing measurable attack success rates and open-source models displaying greater exposure. A baseline detector trained on FENCE achieves 99 percent in-distribution accuracy and maintains strong performance on external benchmarks, underscoring the dataset's robustness for training reliable detection models. FENCE provides a focused resource for advancing multimodal jailbreak detection in finance and for supporting safer, more reliable AI systems in sensitive domains. Warning: This paper includes example data that may be offensive.