AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Attention is Case-Sensitive

Maximilian Dillitzer, Tin Stribor Sohn, Jason Corso, Michael Auerbach

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Training-free token reduction for MLLMs · 1 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
uppercase the target against lowercase context (TE1) (ablation)Qwen2.5-7B-Instruct · Text benchmarksaccuracy: +1.85 ppImproved
alternate the case of the target (TA) (ablation)Qwen2.5-7B-Instruct · Text benchmarksaccuracy: -2.88 ppHurt
uppercase the target against lowercase context (TE1) (ablation)LLaMA-3.1-8B-Instruct · MMLU-Proaccuracy: 29.59 -> 44.39Improved
alternate the case of the target (TA) (ablation)LLaMA-3.1-8B-Instruct · Text benchmarksaccuracy: -13.96 ppHurt
lowercase target with alternating-case context (ADE2) (ablation)non-reasoning LLMs · MMLU-Proaccuracy: -6.84 ppHurt
uppercase the target (TE1) (ablation)Qwen2.5-7B-Instruct · SQuADv2accuracy: +2.13 ppImproved
alternate the case of the target (TA) (ablation)Qwen2.5-7B-Instruct · SQuADv2accuracy: -0.48 ppHurt
uppercase the target (TE) (ablation)Qwen2.5-14B-Instruct · MMLU-Proaccuracy: +9.06 ppImproved
alternate the case of the target (TA2) (ablation)Qwen2.5-7B-Instruct · ARC-Challengeattention mass: 0.061 -> 0.104Improved
apply TA or TE casing to a reasoning model (ablation)Qwen3-4B-Thinking-2507 · Text benchmarksattention mass: +0.23 ppNo clear effect
apply TA or TE casing to a reasoning model (ablation)gpt-oss-20B · Text benchmarksattention mass: +0.07 ppNo clear effect
apply non-standard casing to the text prompt of a VLM (ablation)VLMs (4 models) · RefCOCOgtarget-region attention: +1.55 ppImproved

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