AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

LaVCa: LLM-assisted Visual Cortex Captioning

Takuya Matsuyama, Shinji Nishimoto, Yu Takagi

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 9 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use LaVCa instead of BrainSCUBAcaptioning · NSD sentence-level, subj01prediction accuracy: 0.207 -> 0.246Improved
use LaVCa instead of BrainSCUBAcaptioning · NSD image-level, subj01prediction accuracy: 0.188 -> 0.213Improved
add multiple keywords and Sentence Composer (ablation)LaVCa · NSD sentence-level, subj01prediction accuracy: 0.205 -> 0.246Improved
add Sentence Composer (ablation)LaVCa · subj01 hyperparameter ablationprediction accuracy: 0.230 -> 0.246Improved
use an encoding model to select optimal images (ablation)LaVCa · subj07prediction accuracy: 0.094 -> 0.218Improved
do not shuffle captions within ROI (ablation)LaVCa · OFAprediction accuracy: 0.0280 -> 0.0945Improved
increase optimal images from 5 to 50 (ablation)LaVCa · subj01prediction accuracy: 0.239 -> 0.246Improved
increase optimal images from 50 to 100 (ablation)LaVCa · subj01prediction accuracy: 0.246 -> 0.246No clear effect
increase keywords from 5 to 10 (ablation)LaVCa · subj01prediction accuracy: 0.246 -> 0.241Hurt
swap CLIP-Vision for SigLIP2 (ablation)LaVCa · subj01prediction accuracy: 0.246 -> 0.232Hurt
use gpt-4o instead of Llama3.1-70B for keyword extraction (ablation)LaVCa · subj01prediction accuracy: 0.238 -> 0.246Improved
use ridge regression instead of gradient descent for encoding model (ablation)encoding model · subj01correlation: 0.484 -> 0.500Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy