ViewSpatial-Bench

SenseNova-SI-1.3-InternVL3-8B 61.3% Fading

Models answer 5,712 questions that locate objects from the camera view and from the view of people in the scene. Five task types test egocentric and allocentric reasoning.

Questions were made by automatic generation plus human annotation.

In the authors’ words

Vision-language models (VLMs) have demonstrated remarkable capabilities in understanding and reasoning about visual content, but significant challenges persist in tasks requiring cross-viewpoint understanding and spatial reasoning. We identify a critical limitation: current VLMs excel primarily at egocentric spatial reasoning (from the camera's perspective) but fail to generalize to allocentric viewpoints when required to adopt another entity's spatial frame of reference. We introduce ViewSpatial-Bench, the first comprehensive benchmark designed specifically for multi-viewpoint spatial localization recognition evaluation across five distinct task types, supported by an automated 3D annotation pipeline that generates precise directional labels. Comprehensive evaluation of diverse VLMs on ViewSpatial-Bench reveals a significant performance disparity: models demonstrate reasonable performance on camera-perspective tasks but exhibit reduced accuracy when reasoning from a human viewpoint. By fine-tuning VLMs on our multi-perspective spatial dataset, we achieve an overall performance improvement of 46.24% across tasks, highlighting the efficacy of our approach. Our work establishes a crucial benchmark for spatial intelligence in embodied AI systems and provides empirical evidence that modeling 3D spatial relationships enhances VLMs' corresponding spatial comprehension capabilities.

Abstract of ViewSpatial-Bench: Evaluating Multi-perspective Spatial Localization in Vision-Language Models (arXiv:2505.21500, CC0)

Made by Dingming Li , Yueting Zhuang et al. Released May 2025 Spatial reasoning and counting All tasks public Last in a launch post Jan 2026 , dropped before it saturated Last model added Jan 2026

Level: Everyday Most adults, without training
Scored: Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers: Multiple choice

Paper

Models scored in lab cards

Launch post lists it, from 1 lab

System card reports it, from 1 lab

SenseNova-SI-1.3-InternVL3-8B leads with 61.3%

Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.

Bars Table

1 SenseNova-SI-1.3-InternVL3-8B 61.3%
2 STEP3-VL-10B 46.1%

#	Model	Score	Setting	Date	Source
1	SenseNova-SI-1.3-InternVL3-8B	61.3%		Jan 2026	Model card
2	STEP3-VL-10B	46.1%		Jan 2026	Model card

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card

Labs headline it less often now, last in Jan 2026

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

SenseTime

Date	Launch post	Score	As the post writes it
8 Jan 2026	SenseNova-SI-1.3-InternVL3-8B	61.3	ViewSpatial

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

Date	Document	Score	As the document writes it	Section
14 Jan 2026	STEP3-VL-10B technical report	46.1	ViewSpatial, SeRe (Tables 1 and 3). PaCoRe 48.41	Capabilities

SenseNova-SI-1.3-InternVL3-8B leads with 61.3%

Scores over time

Labs headline it less often now, last in Jan 2026

One system card reports it

More in Vision