PulseAugur
EN
LIVE 11:47:37

New CoTA++ framework combats repetition in dMLLMs

By PulseAugur Editorial · [1 sources] ·

Researchers have introduced CoTA++, a novel framework designed to address the "Repeat Curse" phenomenon in diffusion-based multimodal large language models (dMLLMs). This curse causes dMLLMs to generate repetitive tokens in adjacent positions, degrading response quality. CoTA++ tackles this by re-forming attention routing, implementing a decode-aware refresh mechanism for caching, and using context-token entropy to guide voting. Tested across multiple dMLLMs and caching backends, CoTA++ significantly reduced repetition rates from nearly 100% to under 17%, while recovering accuracy and maintaining caching speedups. AI

IMPACT Addresses a critical failure mode in dMLLMs, potentially improving the reliability and quality of generated content.

RANK_REASON Academic paper detailing a new technical approach to improve LLM performance. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

New CoTA++ framework combats repetition in dMLLMs

How we ranked this

Signal score
9 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Academic paper detailing a new technical approach to improve LLM performance. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Same-day
Cluster formed today. Ranking reflects the current source set at time of score.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Qiyan Zhao, Xiaofeng Zhang, Xu-Yao Zhang, Shijie Han, Zhepeng Li, Shunzhi Zhu, Da-Han Wang ·

    CoTA++: Understanding the Repeat Curse in dMLLMs from an Information Flow Perspective

    arXiv:2610.23290v1 Announce Type: new Abstract: Diffusion-based multimodal large language models (dMLLMs) decode by iterative denoising with bidirectional attention, which precludes the standard autoregressive key--value cache. Practical deployments therefore rely on approximate …

RELATED ENTITIES

RELATED TOPICS