paper-with-me

홈 › Papers

One Token Is Enough: Improving Diffusion Language Models with a Sink Token

2026-01-27 · Zihou Zhang, Zheyong Xie, Li Zhong, Haifeng Liu, Yao Hu, Shaosheng Cao arxiv

Diffusion Language Models (DLMs) have emerged as a compelling alternative to autoregressive approaches, enabling parallel text generation with competitive performance. Despite these advantages, there is a critical instability in DLMs: the moving sink phenomenon. Our analysis indicates that sink tokens exhibit low-norm representations in the Transformer's value space, and that the moving sink phenomenon serves as a protective mechanism in DLMs to prevent excessive information mixing. However, their unpredictable positions across diffusion steps undermine inference robustness. To resolve this, we propose a simple but effective extra sink token implemented via a modified attention mask. Specifically, we introduce a special token constrained to attend solely to itself, while remaining globally visible to all other tokens. Experimental results demonstrate that introducing a single extra token stabilizes attention sinks, substantially improving model performance. Crucially, further analysis confirms that the effectiveness of this token is independent of its position and characterized by negligible semantic content, validating its role as a robust and dedicated structural sink.

📄 PDF Abstract BibTeX arXiv:2601.19657

Code (0)

등록된 구현이 없습니다.

Tasks

Text Generation

Similar Papers 제목 키워드 기반

ASAP: Attention Sink Anchored Pruning

2026-05-21 · Jaehyuk Lee, Hanyoung Kim, Yanggee Kim, Donghun Lee arxiv

Vision Transformers (ViTs) face severe computational bottlenecks due to the quadratic complexity of self-attention at high resolutions. Existing token reduction methods rely on local metrics - such as single-layer attent…

OrthoRank: Token Selection via Sink Token Orthogonality for Efficient LLM inference

2025-07-05 · Seungjun Shin, Jaehoon Oh, Dokwan Oh

Attention mechanisms are central to the success of large language models (LLMs), enabling them to capture intricate token dependencies and implicitly assign importance to each token. Recent studies have revealed the sink…

Sink-Aware Pruning for Diffusion Language Models

2026-02-19 · Aidar Myrzakhan, Tianyi Li, Bowei Guo, Shengkun Tang 외 arxiv

Diffusion Language Models (DLMs) incur high inference cost due to iterative denoising, motivating efficient pruning. Existing pruning heuristics largely inherited from autoregressive (AR) LLMs, typically preserve attenti…

Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing

2026-02-02 · Lingkun Long, Yushi Huang, Shihao Bai, Ruihao Gong 외 arxiv

Diffusion Large Language Models (dLLMs) deliver strong long-context processing capability in a non-autoregressive decoding paradigm. However, the considerable computational cost of bidirectional full attention limits the…

When Graph Tokens Sink: A Mechanistic Analysis of Graph Language Models

2026-06-02 · Ding Zhang, Runtao Zhou, Wenqing Zheng, Rizal Fathony 외 arxiv

Graph Language Models (GLMs) have become a promising direction for adapting Large Language Models (LLMs) to graph learning tasks. By transforming graph topology and node information into graph tokens, GLMs allow LLMs to …

Graph Learning