한양대학교
인공지능연구실

Recent research on the increasing of Large Language Model(LLM)'s reasoning ability has been very successful. LLMs trained with Reinforcement Learning (RL) for reasoning ability show strong performance in challenging tasks like math and coding, even at small model sizes. However, despite the remarkable increase in accuracy of task, the assessment of creativity of LLM's generations has been overlooked in the reasoning task, unlike in the writing task. The main reason why creativity assessment research has not been actively conducted in the reasoning task is that, firstly, it was difficult to define the ‘range of creativity’ and secondly, human evaluation was essential in the process of assessment creativity. To overcome these difficulties, we proposed CLAWS, a novel method that can classify mathematical solutions into ‘Typical, Creative, and Hallucinated’ solutions without human evaluation by using Attention weight by prompt section. CLAWS showed superior performance in five 7-8B math RL models (Deepseek, QWEN, Mathstral, Openmath, OREAL) than five existing white-box detection methods (Perplexity, Logit Entropy, Window Entropy, Hidden Score, Attention Score). We validated this on 4545 math problems from 181 math contests (A(J)HSME, AMC, AIME). The code will be released on github after publication.

Recent advancements in Retrieval-Augmented Generation (RAG) have improved large language models (LLMs) by incorporat- ing external knowledge at inference time. Graph-based RAG systems have emerged as promising approaches, enabling multi-hop reasoning by organizing retrieved information into structured graphs. However, when knowledge graphs are constructed from unstructured documents using LLMs, they often suffer from fragmentation—resulting in disconnected subgraphs that limit inferential coherence and undermine the advantages of graph-based retrieval. To address these limitations, we propose ReGraphRAG, a novel framework designed to reconstruct and enrich fragmented knowledge graphs through three core components: Graph Reorganization, Perspective Expansion, and Query-aware Reranking. Experiments on four benchmarks show that ReGraphRAG outperforms state-of- the-art baselines, achieving over 80% average diversity win rate. Ablation studies highlight the key contributions of graph reorganization and especially perspective expansion to performance gains.

Text-to-image diffusion models have achieved remarkable success in generating high-quality and diverse images. Building on these advancements, diffusion models have also demonstrated exceptional performance in text-guided image editing. A key strategy for effective image editing involves inverting the source image into editable noise maps associated with the target image. However, previous inversion methods face challenges in adhering closely to the target text prompt. The limitation arises because inverted noise maps, while enabling faithful reconstruction of the source image, restrict the flexibility needed for desired edits. To overcome this issue, we propose Editable Noise Map Inversion (ENM Inversion), a novel inversion technique that searches for optimal noise maps to ensure both content preservation and editability. We analyze the properties of noise maps for enhanced editability. Based on this analysis, our method introduces an editable noise refinement that aligns with the desired edits by minimizing the difference between the reconstructed and edited noise maps. Extensive experiments demonstrate that ENM Inversion outperforms existing approaches across a wide range of image editing tasks in both preservation and edit fidelity with target prompts. Our approach can also be easily applied to video editing, enabling temporal consistency and content manipulation across frames.

Methods for mitigating toxic content through masking and infilling often overlook the decision-making process, leading to either insufficient or excessive modifications of toxic tokens. To address this challenge, we propose XDetox, a novel method that integrates token-level toxicity explanations with the masking and infilling detoxification process. We utilized this approach with two strategies to enhance the performance of detoxification. First, identifying toxic tokens to improve the quality of masking. Second, selecting the regenerated sentence by re-ranking the least toxic sentence among candidates. Our experimental results show state-of-the-art performance across four datasets compared to existing detoxification methods. Furthermore, human evaluations indicate that our method outperforms baselines in both fluency and toxicity reduction. These results demonstrate the effectiveness of our method in text detoxification.

Prompt tuning, which freezes all parameters of a pre-trained model and only trains a soft prompt, has emerged as a parameter-efficient approach. For the reason that the prompt initialization becomes sensitive when the model size is small, the prompt transfer that uses the trained prompt as an initialization for the target task has recently been introduced. Since previous works have compared tasks in large categories (e.g., summarization, sentiment analysis), the factors that influence prompt transfer have not been sufficiently explored. In this paper, we characterize the question answering task based on features such as answer format and empirically investigate the transferability of soft prompts for the first time. We analyze the impact of initialization during prompt transfer and find that the train dataset size of source and target tasks have the influence significantly. Furthermore, we propose a novel approach for measuring catastrophic forgetting and investigate how it occurs in terms of the amount of evidence. Our findings can help deeply understand transfer learning in prompt tuning.

Attention mechanisms have become a crucial aspect of deep learning, particularly in natural language processing (NLP) tasks. However, in tasks such as constituency parsing, attention mechanisms can lack the directional information needed to form sentence spans. To address this issue, we propose a Bidirectional masked and N-gram span Attention (BNA) model, which is designed by modifying the attention mechanisms to capture the explicit dependencies between each word and enhance the representation of the output span vectors. The proposed model achieves state-of-the-art performance on the Penn Treebank and Chinese Penn Treebank datasets, with F1 scores of 96.47 and 94.15, respectively. Ablation studies and analysis show that our proposed BNA model effectively captures sentence structure by contextualizing each word in a sentence through bidirectional dependencies and enhancing span representation.

An important problem of the sequence-to-sequence neural models widely used in abstractive summarization is exposure bias. To alleviate this problem, re-ranking systems have been applied in recent years. Despite some performance improvements, this approach remains underexplored. Previous works have mostly specified the rank through the ROUGE score and aligned candidate summaries, but there can be quite a large gap between the lexical overlap metric and semantic similarity. In this paper, we propose a novel training method in which a re-ranker balances the lexical and semantic quality. We further newly define false positives in ranking and present a strategy to reduce their influence. Experiments on the CNN/DailyMail and XSum datasets show that our method can estimate the meaning of summaries without seriously degrading the lexical aspect. More specifically, it achieves an 89.67 BERTScore on the CNN/DailyMail dataset, reaching new state-of-the-art performance.