AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Evaluating and Enhancing Negation Comprehension in Remote Sensing MLLMs

Haochen Han, Jue Wang, Alex Jinpeng Wang, Fangming Liu

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Benchmarks and analysis of 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
apply existing TTA methods (Tent, SAR, TLM) to MLLMs on negation queriesQwen2.5-VL-7B · RS-Neg VQA (total)accuracy: 74.96 -> 73.09 (Tent), 65.21 (SAR), 69.30 (TLM)Hurt
add NeFo test-time adaptationQwen2.5-VL-7B · RS-Neg VQA (total)accuracy: 74.96 -> 79.52Improved
add NeFo test-time adaptationQwen2.5-VL-7B · RS-Neg MCQ (total)accuracy: 54.29 -> 65.46Improved
add NeFo test-time adaptationRS-LLaVA-7B · RS-Neg MCQ (total)accuracy: 23.72 -> 24.46No clear effect
add truth-value inversion loss (ablation)NeFo (Qwen2.5-VL-7B) · RS-Neg VQA (total, ablation)accuracy: 74.05 -> 79.52Improved
add knowledge retaining loss (ablation)NeFo (Qwen2.5-VL-7B) · RS-Neg MCQ (total, ablation)accuracy: 22.71 -> 65.46Improved
add entropy minimization on negation-masked queries (ablation)NeFo (Qwen2.5-VL-7B) · RS-Neg MCQ (total, ablation)accuracy: 29.06 -> 65.46Improved
add entropy minimization on negation-masked queries (ablation)NeFo (Qwen2.5-VL-7B) · RS-Neg VQA (total, ablation)accuracy: 76.84 -> 79.52Improved
zero-shot transfer of MCQ-adapted NeFo to RS-Neg groundingQwen2.5-VL-7B · RS-Neg Grounding IoU@0.5 (zero-shot)accuracy: 54.03 -> 63.37Improved
zero-shot transfer of MCQ-adapted NeFo to RS-Neg groundingGeoReason-7B · RS-Neg Grounding IoU@0.5 (zero-shot)accuracy: 46.94 -> 45.85Hurt
zero-shot transfer of NeFo to FloodNet VQAQwen2.5-VL-7B · FloodNet VQA (zero-shot, total)accuracy: 37.32 -> 57.47Improved
apply NeFo across Qwen3-VL scales 2B-32B (ablation)Qwen3-VL · RS-Neg VQA (total)accuracy: 64.61 -> 74.51 (2B)Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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