AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

The Language of Visual Attention: Modeling Scanpaths via Autoregressive Token Prediction

Susmit Agrawal, Matthias Bethge, Matthias Kuemmerer

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis

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
fine-tune a VLM for autoregressive coordinate-token scanpath prediction instead of DeepGaze IIIDeepGaze3.5-VL · MIT1003Information Gain: 1.49 -> 2.18Improved
give DeepGaze III the larger InternViT-6B encoder (ablation)DeepGaze III · MIT1003Information Gain: 1.45 -> 1.49Improved
add duration conditioning to the tokens (ablation)DeepGaze3.5-VL · MIT1003Information Gain: 2.03 -> 2.12Improved
add duration conditioning to the tokens (ablation)DeepGaze3.5-VL · COCO-FreeViewInformation Gain: 2.15 -> 2.30Improved
add duration conditioning to the tokens (ablation)DeepGaze3.5-VL · DAEMONSInformation Gain: 2.67 -> 2.84Improved
add subject identity token to the prompt (ablation)DeepGaze3.5-VL · MIT1003 (leave-subject-out)Information Gain: +0.09 bits/fixImproved
put the search target in the promptDeepGaze3.5-VL · COCO-Search18Information Gain: 2.15 -> 3.27Improved
search conditioning when target is absent (ablation)DeepGaze3.5-VL · COCO-Search18 (target absent)Information Gain: 2.29 vs 2.15 (free viewing)No clear effect
train on all five datasets combined instead of in-distribution only (ablation)DeepGaze3.5-VL · MIT1003Information Gain: 2.03 -> 2.18Improved
train on all datasets including the target vs leave-one-dataset-out (ablation)DeepGaze3.5-VL · MIT1003Information Gain: 1.94 -> 2.18Improved
stronger multimodal backbone (higher MMMU) (ablation)LoRA-tuned VLMs · MIT1003Information Gain: Spearman rho = 0.93Improved
scale SmolVLM2 backbone from 256M to 2.2B (ablation)SmolVLM2 · MIT1003Information Gain: 0.85 bits across scalesNo clear effect

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