# MindEdit-Bench: Benchmarking Object-Level Counterfactual Spatial Reasoning in VLMs from In-the-Wild Photos

arXiv: https://arxiv.org/abs/2607.00491 · HTML: https://mlsanity.com/paper/2607.00491/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 16 of its 20 references that are in the map.
- First posted on arXiv 1 Jul 2026.

## Placed in

- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)
- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 16 of 16)

- [ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes](https://mlsanity.com/paper/1702.04405/) (2017)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [SAM 3: Segment Anything with Concepts](https://mlsanity.com/paper/2511.16719/) (2025)
- [SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities](https://mlsanity.com/paper/2401.12168/) (2024)
- [Depth Anything 3: Recovering the Visual Space from Any Views](https://mlsanity.com/paper/2511.10647/) (2025)
- [Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces](https://mlsanity.com/paper/2412.14171/) (2024)
- [Kimi K2.5: Visual Agentic Intelligence](https://mlsanity.com/paper/2602.02276/) (2026)
- [SQA3D: Situated Question Answering in 3D Scenes](https://mlsanity.com/paper/2210.07474/) (2022)
- [MMSI-Bench: A Benchmark for Multi-Image Spatial Intelligence](https://mlsanity.com/paper/2505.23764/) (2025)
- [ViewSpatial-Bench: Evaluating Multi-perspective Spatial Localization in Vision-Language Models](https://mlsanity.com/paper/2505.21500/) (2025)
- [MindCube: Spatial Mental Modeling from Limited Views](https://mlsanity.com/paper/2506.21458/) (2025)
- [Scaling Spatial Intelligence with Multimodal Foundation Models](https://mlsanity.com/paper/2511.13719/) (2025)
- [SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence](https://mlsanity.com/paper/2505.17012/) (2025)
- [SpatialTree: How Spatial Abilities Branch Out in MLLMs](https://mlsanity.com/paper/2512.20617/) (2025)
- [SpatialEdit: Benchmarking Fine-Grained Image Spatial Editing](https://mlsanity.com/paper/2604.04911/) (2026)
- [Holi-Spatial: Evolving Video Streams into Holistic 3D Spatial Intelligence](https://mlsanity.com/paper/2603.07660/) (2026)

## Built on by (most cited first, 4 of 4)

- [CrossProjection: Geometric Grounding Beyond Viewpoint Change in Architectural Drawings](https://mlsanity.com/paper/2608.00473/) (2026)
- [When Do Frozen VLMs Respond to Image-Free Object-Token Edits? An Answer-Key-Free Protocol and What It Reveals](https://mlsanity.com/paper/2609.03429/) (2026)
- [GaugeVLM: Structuring Spatial Supervision with Measured Geometric Interventions](https://mlsanity.com/paper/2609.38285/) (2026)
- [Frame Evidence:A Benchmark for Reference Frames and Visual Evidence in Multi-Image Spatial Reasoning](https://mlsanity.com/paper/2610.19564/) (2026)

## Abstract (arXiv)

Benchmarks for vision-language models (VLMs) mostly test observational spatial reasoning: models describe relations already visible in the input. Existing what-if tasks typically vary the observer while keeping the scene fixed. Can VLMs instead predict the consequences of hypothetically moving or rotating an object? We introduce MindEdit-Bench, a benchmark of six spatial reasoning tasks built from three-photo smartphone triplets of newly captured indoor scenes via an automatic in-the-wild 3D scene-graph extraction pipeline. Four tasks probe perception and perspective transformation over observed structure; two new tasks, L4 (spatial editing) and L5 (cross-view visibility editing), probe object-level counterfactual reasoning, where correct answers are absent from all input images. Each question provides 8-24 structured answer choices, enabling answer-letter-level diagnosis of spatial and fallback errors. The benchmark covers 120 private indoor scenes not drawn from public datasets, reducing public-data pretraining-overlap risk. Across 15 VLMs on 1,003 human-verified questions, task-wise mean VLM accuracy is only 8%-31%, versus 81%-97% human majority-vote accuracy. The pooled human--best-VLM gap is 53 pp, with at least 39 pp on every task. The structured answer space further reveals non-uniform failures, including weaker camera-depth-axis inference and fallback behavior on difficult visibility-editing cases.