Injecting pre-computed layout priors from RT-DETR into VLM prompts raises markdown F1 from 0.37 to 0.92 on a 10k-page OOD benchmark and cuts infinite-loop failures across domains.
Title resolution pending
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 3roles
background 1polarities
background 1representative citing papers
A lightweight Q-Former proxy trained on VLM hidden states reveals that localization signals peak in input-dependent intermediate layers, not the final layers used by standard editing pipelines.
The Gluon Condensation model fits gamma-ray spectra of TXS 0506+056 and NGC 1068 with neutrino predictions consistent with IceCube, but predicts unphysical hardening for SNR G54.1+0.3.
citing papers explorer
-
Structured Layout Priors for Robust Out-of-Distribution Visual Document Understanding
Injecting pre-computed layout priors from RT-DETR into VLM prompts raises markdown F1 from 0.37 to 0.92 on a 10k-page OOD benchmark and cuts infinite-loop failures across domains.
-
Analysis-by-Proxy: Localization Signals in VLMs Operating as Condition Encoders
A lightweight Q-Former proxy trained on VLM hidden states reveals that localization signals peak in input-dependent intermediate layers, not the final layers used by standard editing pipelines.
-
A Unified Explanation of Gamma-Ray and Neutrino Spectra from Astrophysical Sources Based on the Gluon Condensation Model
The Gluon Condensation model fits gamma-ray spectra of TXS 0506+056 and NGC 1068 with neutrino predictions consistent with IceCube, but predicts unphysical hardening for SNR G54.1+0.3.