1 citations · 1 across the 2 of their papers we have counts for
4 papers
Readout is not Recovery: Dissociating Coordinate Emission from Visual-Corruption Repair in Vision-Language Models
Drandreb Earl Juanico
VLM bounding-box localization is both language generation and spatial commitment. Parseable fields such as bbox_2d make localization easy to score, but dimensions that emit coordin…
FLIP: Final Layer Inference-Time Probing for Vision-Language Models
Drandreb Earl O. Juanico, Rowel O. Atienza
We present FLIP, a final-layer inference-time probe for testing whether a logit-facing intervention site in an open-weight vision-language model (VLM) supports structured, task-lin…
Interpretable Open-Vocabulary Referring Object Detection with Reverse Contrast Attention
Drandreb Earl O. Juanico, Rowel O. Atienza, Jeffrey Kenneth Go
We propose Reverse Contrast Attention (RCA), a plug-in method that enhances object localization in vision-language transformers without retraining. RCA reweights final-layer attent…
Simulated effects of site salinity and inundation on long-term growth trajectory and carbon sequestration in monospecific plantation in the Philippines
Drandreb Earl Juanico, Severino Salmo
A mathematical model of coastal forest growth is proposed to describe and test the effects of salinity and inundation in the long-term growth performance and carbon sequestration o…