paper-with-me

홈 › Papers

What Gets Unmasked First? Trajectory Analysis of Diffusion Models for Graph-to-Text Generation

2026-05-29 · Qing Wang, Jacob Devasier, Chengkai Li arxiv

We present the first systematic study of masked diffusion language models (MDLMs) for graph-to-text generation. We analyze MDLM generation trajectories -- the order in which tokens are unmasked during iterative decoding -- and find that, unlike autoregressive LLMs which generate text linearly, MDLMs naturally prioritize entities first, followed by relational and function words, with structural tokens resolved last. We further identify a previously undocumented failure mode of supervised fine-tuning: SFT disrupts this strategy by prematurely anchoring structural sentence-ending tokens early in the decoding trajectory, effectively fixing the output length which can lead to omitted or hallucinated information. To address this, we propose lambda-scaled structural decoding, a training-free inference-time modification that downweights structural token confidence and recovers +9.4 BLEU-4. Finally, we introduce Graph-LLaDA, which integrates a Graph Transformer encoder into LLaDA's decoding process to explicitly incorporate relational graph structure. Cross-dataset evaluation on LAGRANGE reveals that previous baselines overfit to dataset-specific patterns, while LLM- and MDLM-based approaches generalize significantly better.

📄 PDF Abstract BibTeX arXiv:2605.31564

Code (0)

등록된 구현이 없습니다.

Tasks

Text Generation

Similar Papers 제목 키워드 기반

M$^3$PC: Test-time Model Predictive Control for Pretrained Masked Trajectory Model

2024-12-07 · Kehan Wen, Yutong Hu, Yao Mu, Lei Ke

Recent work in Offline Reinforcement Learning (RL) has shown that a unified Transformer trained under a masked auto-encoding objective can effectively capture the relationships between different modalities (e.g., states,…

D4RLmodelModel Predictive ControlReinforcement Learning (RL)

Bootstrapped Masked Autoencoders for Vision BERT Pretraining

2022-07-14 · Xiaoyi Dong, Jianmin Bao, Ting Zhang, Dongdong Chen 외

We propose bootstrapped masked autoencoders (BootMAE), a new approach for vision BERT pretraining. BootMAE improves the original masked autoencoders (MAE) with two core designs: 1) momentum encoder that provides online f…

DecoderObject DetectionPredictionSelf-Supervised Image Classification+1

Remote Pulse Estimation in the Presence of Face Masks

2021-01-11 · Jeremy Speth, Nathan Vance, Patrick Flynn, Kevin Bowyer 외

Remote photoplethysmography (rPPG), a family of techniques for monitoring blood volume changes, may be especially useful for widespread contactless health monitoring using face video from consumer-grade visible-light cam…

Data AugmentationHeart rate estimation

Nerve Block Target Localization and Needle Guidance for Autonomous Robotic Ultrasound Guided Regional Anesthesia

2023-08-07 · Abhishek Tyagi, Abhay Tyagi, Manpreet Kaur, Richa Aggarwal 외

Visual servoing for the development of autonomous robotic systems capable of administering UltraSound (US) guided regional anesthesia requires real-time segmentation of nerves, needle tip localization and needle trajecto…

Anatomyobject-detectionObject DetectionObject Tracking+1

Exploring Prediction Targets in Masked Pre-Training for Speech Foundation Models

2024-09-16 · Li-Wei Chen, Takuya Higuchi, He Bai, Ahmed Hussen Abdelaziz 외

Speech foundation models, such as HuBERT and its variants, are pre-trained on large amounts of unlabeled speech data and then used for a range of downstream tasks. These models use a masked prediction objective, where th…

DenoisingPrediction