AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

EgoNight: Towards Egocentric Vision Understanding at Night with a Challenging Benchmark

Deheng Zhang, Yuqian Fu, Runyi Yang, Yang Miao, Tianwen Qian, Xu Zheng, Guolei Sun, Ajad Chhatkuli, Xuanjing Huang, Yu-Gang Jiang, Luc Van Gool, Danda Pani Paudel

ICLR 2026 · Vision-language and multimodal › Visual question answering · approach: Visual reasoning diagnostics · 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
evaluate nighttime vs daytime for egocentric MLLMsMLLMs (averaged) · EgoNight-SyntheticVQA accuracy: -32.8%Hurt
scale Qwen2.5-VL from 7B to 72BQwen2.5-VL · EgoNight-VQA avgaccuracy (%): 13.44 -> 18.99Improved
full fine-tuning on EgoNight train split vs zero-shot (ablation)Qwen2.5-VL-7B · EgoNight realaccuracy (%): 16.40 -> 25.61Improved
tune LLM only vs vision encoder only (ablation)Qwen2.5-VL-7B · EgoNight syntheticaccuracy (%): 29.74 -> 35.50Improved
full fine-tuning vs LLM-only tuning (ablation)Qwen2.5-VL-7B · EgoNight syntheticaccuracy (%): 35.50 -> 36.25Improved
fine-tune on synthetic nighttime data only (ablation)Qwen2.5-VL-7B · EgoNight realaccuracy (%): 14.83 -> 20.57Improved
fine-tune (LLM-only) on reasoning tasks (navigation) (ablation)Qwen2.5-VL-7B · EgoNight navigationaccuracy (%): 17.870 -> 19.918Improved
fine-tune (LLM-only) on perception tasks (object recognition) (ablation)Qwen2.5-VL-7B · EgoNight object recognitionaccuracy (%): 8.435 -> 35.855Improved
use night query instead of day query for retrievalGPT-4.1 · EgoNight-Synthetic spatial retrievalAcc R@1 (%): 75.6 -> 54.1Hurt
MLLM (GPT-4.1) vs feature-based retrieval for temporal localizationGPT-4.1 · EgoNight-Sofia Night->DaymIoU (%): 33.4 -> 15.5Hurt
use fisheye-specific depth model (UniK3D) vs general Depth Anythingdepth estimation · EgoNight-Synthetic nightAbs Rel: 0.302 -> 0.253Improved

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