analiza aiqbench
EchoChange introduces a discrete diffusion language model that replaces traditional autoregressive decoding with iterative masked-token denoising, specifically targeting the reduction of cascading factual errors in remote-sensing disaster analysis. By employing draft-aware dual-pass training and confidence-guided remasking, this approach allows the system to revise uncertain content before finalizing descriptions, offering a significant improvement over existing baselines on the RSCC benchmark. This work represents an important advancement in applying generative AI models for high-stakes visual reasoning tasks where factual accuracy is paramount.
EchoChange: A Diffusion Language Model with Dual Pass Remasking for Factual Remote Sensing Disaster Change Captioning
arXiv:2608.01856v1 Announce Type: new Abstract: Bi-temporal remote-sensing disaster change captioning often needs to identify sparse and spatially localized changes across large pre- and post-event scenes and then translate them into coherent, factual descriptions. However, existing change captioning methods always follow an autoregressive decoding paradigm to generate the change description and thus an early misinterpretation of the changed object, event, or spatial relation becomes an irreversible premise for subsequent text, amplifying visual ambiguity into cascading factual errors. To address this limitation, we propose EchoChange, a multimodal discrete diffusion language model that formulates change captioning as iterative masked-token denoising rather than left-to-right generation. By repeatedly revising the entire caption while conditioning on the image pair, EchoChange can reconsider uncertain content and correct imperfect intermediate predictions. We further introduce draft-aware dual-pass training, a progressive masking curriculum, and confidence-guided remasking to align training with iterative inference. Extensive experiments on the RSCC benchmark show that EchoChange substantially outperforms both general-purpose and remote-sensing-specific baselines across lexical and semantic metrics. The EchoChange Project is at https://github.com/sundongwei/EchoChange_Project
Abstract:Bi-temporal remote-sensing disaster change captioning often needs to identify sparse and spatially localized changes across large pre- and post-event scenes and then translate them into coherent, factual descriptions. However, existing change captioning methods always follow an autoregressive decoding paradigm to generate the change description and thus an early misinterpretation of the changed object, event, or spatial relation becomes an irreversible premise for subsequent text, amplifying visual ambiguity into cascading factual errors. To address this limitation, we propose EchoChange, a multimodal discrete diffusion language model that formulates change captioning as iterative masked-token denoising rather than left-to-right generation. By repeatedly revising the entire caption while conditioning on the image pair, EchoChange can reconsider uncertain content and correct imperfect intermediate predictions. We further introduce draft-aware dual-pass training, a progressive masking curriculum, and confidence-guided remasking to align training with iterative inference. Extensive experiments on the RSCC benchmark show that EchoChange substantially outperforms both general-purpose and remote-sensing-specific baselines across lexical and semantic metrics. The EchoChange Project is at this https URL
Artificial Intelligence (cs.AI)
arXiv:2608.01856 [cs.AI]
(or arXiv:2608.01856v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.01856 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 260 cuvinte
Subiecte conexe
NAVIRA: Decoupled Stochastic Remasking for Masked Diffusion Language Models
arXiv · 5 iun. 2026
CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models
arXiv · 8 oct. 2026
Parallelism, critical windows, and separations among diffusion language models
arXiv · 18 sept. 2026
D^2SD: Accelerating Speculative Decoding with Dual Diffusion Draft Models
arXiv · 5 iun. 2026