Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Unified Energy for Invariant and Independent Decoding in Diffusion Language Models

About

Diffusion Language Models (DLMs) enable parallel text generation by iteratively denoising a full sequence, offering attractive flexibility compared to auto-regressive (AR) decoding. However, existing methods fail to fully capture token relationships, leading to a performance gap relative to AR baselines, especially as the degree of parallelism increases. In this paper, we give a systematic analysis of the gap, identifying three key factors: (i) model capacity, (ii) dependency, and (iii) invariance. To address these issues, we first propose an invariant energy (Inv-E) together with an effective sampling-based estimator to handle the invariance issue. By further combining with the independent energy (Ind-E), we obtain a unified energy (Uni-E), that accounts for all these factors. Uni-E enjoys a unique advantage: it can be computed exactly without sampling-based partition estimation. Besides, Uni-E is model agnostic and can therefore be scaled to models of arbitrary size. We further prove that Uni-E can correct the distribution shift caused by dependency and invariance. Extensive experiments across Diffusion Language Models (DLMs) and Diffusion Large Language Models (DLLMs) demonstrate the effectiveness of the proposed Uni-E.

Yuchen Yan, Minkai Xu, Zaiquan Yang, Yatao Bian• 2026

Related benchmarks

TaskDatasetResultRank
Commonsense ReasoningHellaSwag
HellaSwag Accuracy73.7
897
Mathematical ReasoningMATH 500
Accuracy39.2
589
Multitask Language UnderstandingMMLU
Accuracy66.5
568
Language ModelingOpenWebText
Perplexity16.54
190
Multi-task Language UnderstandingMMLU
Top-1 Accuracy66.5
46
Language ModelingLAMBADA zero-shot (test)--
44
Language GenerationLanguage Modeling Evaluation Set
Generative Perplexity (Llama2)13.1
42
ReasoningLiveBench
Accuracy10.9
40
Language ModelingPTB zero-shot
Perplexity86.18
35
Language ModelingPubmed zero-shot
Perplexity41.35
30
Showing 10 of 17 rows

Other info

Follow for update