Injin Kong

Injin Kong, Hyoungjoon Lee, Yohan Jo

Post-training pretrained Autoregressive models (ARMs) into Masked Diffusion models (MDMs) has emerged as a cost-effective strategy to overcome the limitations of sequential generation. However, the internal algorithmic transformations induced by this paradigm shift remain unexplored, leaving it unclear whether post-trained MDMs acquire genuine bidirectional reasoning capabilities or merely repackage autoregressive heuristics. In this work, we address this question by conducting a comparative circuit analysis of ARMs and their MDM counterparts. Our analysis reveals a systematic "mechanism shift" dependent on the structural nature of the task. Structurally, we observe a distinct divergence: while MDMs largely retain autoregressive circuitry for tasks dominated by local causal dependencies, they abandon initialized pathways for global planning tasks, exhibiting distinct rewiring characterized by increased early-layer processing. Semantically, we identify a transition from sharp, localized specialization in ARMs to distributed integration in MDMs. Through these findings, we conclude that diffusion post-training does not merely adapt model parameters but fundamentally reorganizes internal computation to support non-sequential global planning.

Injin Kong, Hyoungjoon Lee, Yohan Jo

Continuous diffusion language models lag behind autoregressive transformers, partly because diffusion is applied in spaces poorly suited to language denoising and token recovery. We propose DiHAL, a geometry-guided diffusion-transformer hybrid that asks where diffusion should enter a pretrained transformer. DiHAL scores layers with geometry-based proxies, selects a diffusion-friendly hidden-state interface, and replaces the lower transformer prefix with a diffusion bridge while retaining the upper layers and original LM head. By reconstructing the selected-layer hidden state rather than tokens, DiHAL avoids direct continuous-to-discrete recovery. Experiments on 8B-scale backbones show that the geometry score predicts effective shallow insertion layers under a fixed bridge-training protocol and that hidden-state recovery improves over continuous diffusion baselines in a diagnostic comparison matching the diffusion/recovery training budget. These results suggest that hidden-state geometry helps identify where diffusion-based replacement is feasible inside pretrained language models.

Naeun Lee, Hyunjong Kim, Sunghwan Choi et al.

Despite strong performance of Multimodal Large Language Models (MLLMs) on multimodal tasks, predicting whether and why an image is persuasive remains challenging. We first show that prompting MLLMs to reason before prediction does not consistently help, and can even reduce persuasiveness prediction performance, suggesting that naively generated rationales are unreliable signals for this task. Yet, no established methodology exists for training MLLMs to reason about visual persuasion or evaluating whether their rationales faithfully support their decisions. To address this gap, we show empirically and theoretically that diverse teacher-generated rationales, when used for supervised fine-tuning, improve visual persuasiveness prediction. We further introduce a three-dimensional faithfulness evaluation framework covering rationale-to-decision consistency, rationale-to-image groundedness, and rationale-to-decision sensitivity. Applying this framework shows that prediction performance alone does not guarantee faithful rationales, while rationale-to-decision sensitivity is most aligned with human rationale preferences. These findings motivate faithfulness-aware training objectives and scalable rationale supervision for visual persuasiveness evaluation. Our code and dataset will be made publicly available.

Jongwook Han, Jongwon Lim, Injin Kong et al.

Large language models (LLMs) can express different values in two distinct ways: (1) intrinsic expression, reflecting the model's inherent values learned during training, and (2) prompted expression, elicited by explicit prompts. Given their widespread use in value alignment and persona steering, it is paramount to clearly understand their underlying mechanisms, particularly whether they mostly overlap (as one might expect) or rely on substantially different mechanisms, but this remains largely understudied. We analyze this at the mechanistic level using two approaches: (1) value vectors, feature directions representing value mechanisms extracted from the residual stream, and (2) value neurons, MLP neurons that contribute to value expressions. We demonstrate that intrinsic and prompted value mechanisms partly share common components that are crucial for inducing value expression, but also possess unique elements that manifest in different ways. As a result, these mechanisms lead to different degrees of value steerability (prompted > intrinsic) and response diversity (intrinsic > prompted). In particular, components unique to the intrinsic mechanism seem to promote lexical diversity in responses, whereas those specific to the prompted mechanism primarily strengthen instruction following, taking effect even in distant tasks like jailbreaking.

InJin Kong, Shinyee Kang, Yuna Park et al.

This study investigates the stylistic differences among various Bible translations using a Variational Autoencoder (VAE) model. By embedding textual data into high-dimensional vectors, the study aims to detect and analyze stylistic variations between translations, with a specific focus on distinguishing the American Standard Version (ASV) from other translations. The results demonstrate that each translation exhibits a unique stylistic distribution, which can be effectively identified using the VAE model. These findings suggest that the VAE model is proficient in capturing and differentiating textual styles, although it is primarily optimized for distinguishing a single style. The study highlights the model's potential for broader applications in AI-based text generation and stylistic analysis, while also acknowledging the need for further model refinement to address the complexity of multi-dimensional stylistic relationships. Future research could extend this methodology to other text domains, offering deeper insights into the stylistic features embedded within various types of textual data.