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

• URL: http://arxiv.org/abs/2605.31564v1
• Date: Fri, 29 May 2026 17:29:35 GMT
• Title: What Gets Unmasked First? Trajectory Analysis of Diffusion Models for Graph-to-Text Generation
• Abstract summary: We present the first systematic study of masked diffusion language models for graph-to-text generation.<n>We find that MDLMs naturally prioritize entities first, followed by relational and function words, with structural tokens resolved last.<n>We introduce Graph-LLaDA, which integrates a Graph Transformer encoder into LLaDA's decoding process to explicitly incorporate graph structure.
• Score: 7.344034836280294
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• Weave of Formal Thought [51.56484100374058]
  We introduce Weave of Formal Thought (WoFT), a paradigm uniting rigorous syntactic validation with learned structural representations.<n>We present a latent-variable fine-tuning method training the language model to interleave non-terminal grammar symbols directly into generation.<n>For Python, fine-tuning StarCoder2-3B with our RWS objective reduces per-token cross-entropy by 14.3% relative to a text-only SFT baseline.
  arXiv  Detail & Related papers  (2026-06-24T15:58:11Z)
• Teaching LLMs to See Graphs: Unifying Text and Structural Reasoning [0.0]
  We introduce the Graph Transformer Language Model (GTLM), a novel architecture that enables pretrained LLMs to process graph topologies.<n>GTLM is exceptionally parameter-efficient, injecting graph-aware attention biases directly into the LLM's attention modules.<n>We demonstrate that GTLM attention heads implicitly learn to simulate message passing, explaining its superior performance on algorithmic tasks.
  arXiv  Detail & Related papers  (2026-05-11T09:19:55Z)
• Toward Graph-Tokenizing Large Language Models with Reconstructive Graph Instruction Tuning [17.712367049197212]
  Key challenge is to align graph data with language spaces so that large language models (LLMs) can better comprehend graphs.<n>GTokenLLMs encode complex structures and lengthy texts into a graph token sequence, and then align them with text tokens via language instructions tuning.<n>Despite their initial success, our information-theoretic analysis reveals that existing GTokenLLMs rely solely on text supervision from language instructions.
  arXiv  Detail & Related papers  (2026-03-02T02:26:54Z)
• SAFT: Structure-Aware Fine-Tuning of LLMs for AMR-to-Text Generation [50.277959544420455]
  SAFT is a structure-aware fine-tuning approach that injects graph topology into pretrained language models.<n>We compute direction-sensitive positional encodings from the magnetic Laplacian of transformed AMRs.<n> SAFT sets a new state-of-the-art on AMR 3.0 with a 3.5 BLEU improvement over baselines.
  arXiv  Detail & Related papers  (2025-07-15T18:12:57Z)
• Compile Scene Graphs with Reinforcement Learning [69.36723767339001]
  Next-token prediction is the fundamental principle for training large language models (LLMs)<n>We introduce R1-SGG, a multimodal LLM (M-LLM) initially trained via supervised fine-tuning (SFT) on the scene graph dataset.<n>We design a set of graph-centric rewards, including three recall-based variants -- Hard Recall, Hard Recall+Relax, and Soft Recall.
  arXiv  Detail & Related papers  (2025-04-18T10:46:22Z)
• GL-Fusion: Rethinking the Combination of Graph Neural Network and Large Language model [63.774726052837266]
  We introduce a new architecture that deeply integrates Graph Neural Networks (GNNs) with Large Language Models (LLMs)<n>We introduce three key innovations: (1) Structure-Aware Transformers, which incorporate GNN's message-passing capabilities directly into LLM's transformer layers; (2) Graph-Text Cross-Attention, which processes full, uncompressed text from graph nodes and edges; and (3) GNN-LLM Twin Predictor, enabling LLM's flexible autoregressive generation alongside GNN's scalable one-pass prediction.
  arXiv  Detail & Related papers  (2024-12-08T05:49:58Z)
• CodecLM: Aligning Language Models with Tailored Synthetic Data [51.59223474427153]
  We introduce CodecLM, a framework for adaptively generating high-quality synthetic data for instruction-following abilities. We first encode seed instructions into metadata, which are concise keywords generated on-the-fly to capture the target instruction distribution. We also introduce Self-Rubrics and Contrastive Filtering during decoding to tailor data-efficient samples.
  arXiv  Detail & Related papers  (2024-04-08T21:15:36Z)
• Large Language Models as Topological Structure Enhancers for Text-Attributed Graphs [3.5627549694751184]
  Large language models (LLMs) have revolutionized the field of natural language processing (NLP)<n>This work explores how to leverage the information retrieval and text generation capabilities of LLMs to refine/enhance the topological structure of text-attributed graphs (TAGs) under the node classification setting.
  arXiv  Detail & Related papers  (2023-11-24T07:53:48Z)
• Structural Adapters in Pretrained Language Models for AMR-to-text Generation [59.50420985074769]
  Previous work on text generation from graph-structured data relies on pretrained language models (PLMs) We propose StructAdapt, an adapter method to encode graph structure into PLMs.
  arXiv  Detail & Related papers  (2021-03-16T15:06:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.