technique

lambda-scaled structural decoding

techniqueactiveprovisionallambda-scaled-structural-decoding-8842df8e·1 events·first seen 16d ago

Aliases: lambda-scaled structural decoding

Co-occurring entities

BLEU-4 Graph Transformer Diffusion Language Models Graph-LLaDA LLaDA LAGRANGE

More like this (12)

blockwise decoding Grammar-Constrained Decoding DFM Decoder Scaling Laws for Neural Language Models intra-frame entropy-guided sparsification speculative decoding flow-matching decoder Parallel Box Decoding code synthesis LLMs SDE (Stochastic Differential Equation LR scaling)Scaling LLM Reasoning from Minimal Labels: A Semi-Supervised Framework with a Lightweight Verifier Continual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMs

Recent events (1)

6arXiv · cs.CL·16d ago·source ↗

Trajectory Analysis of Masked Diffusion LMs for Graph-to-Text Generation with Lambda-Scaled Structural Decoding

This paper presents the first systematic study of masked diffusion language models (MDLMs) for graph-to-text generation, analyzing the order in which tokens are unmasked during iterative decoding. The authors find MDLMs naturally unmask entities first, then relational/function words, then structural tokens—a pattern disrupted by supervised fine-tuning, which prematurely anchors structural tokens and causes hallucination or omission. They propose lambda-scaled structural decoding, a training-free inference-time fix that recovers +9.4 BLEU-4, and introduce Graph-LLaDA, which integrates a Graph Transformer encoder into LLaDA's decoding process. Cross-dataset evaluation on the LAGRANGE benchmark shows prior baselines overfit to dataset-specific patterns while MDLM-based approaches generalize better.

Frontier Model Releases Evaluation and Benchmarking BLEU-4 Graph Transformer Diffusion Language Models +5 more