# On Geometric Understanding and Learned Priors in Feed-forward 3D Reconstruction Models

arXiv: https://arxiv.org/abs/2512.11508 · HTML: https://mlsanity.com/paper/2512.11508/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 29 of its 44 references that are in the map.
- First posted on arXiv 12 Dec 2025.

## Builds on (most cited first, 29 of 29)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields](https://mlsanity.com/paper/2111.12077/) (2021)
- [DUSt3R: Geometric 3D Vision Made Easy](https://mlsanity.com/paper/2312.14132/) (2023)
- [VGGT: Visual Geometry Grounded Transformer](https://mlsanity.com/paper/2503.11651/) (2025)
- [SuperGlue: Learning Feature Matching with Graph Neural Networks](https://mlsanity.com/paper/1911.11763/) (2019)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [LoFTR: Detector-Free Local Feature Matching with Transformers](https://mlsanity.com/paper/2104.00680/) (2021)
- [Grounding Image Matching in 3D with MASt3R](https://mlsanity.com/paper/2406.09756/) (2024)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [iBOT: Image BERT Pre-Training with Online Tokenizer](https://mlsanity.com/paper/2111.07832/) (2021)
- [MVSNet: Depth Inference for Unstructured Multi-view Stereo](https://mlsanity.com/paper/1804.02505/) (2018)
- [Network Dissection: Quantifying Interpretability of Deep Visual Representations](https://mlsanity.com/paper/1704.05796/) (2017)
- [Depth Anything 3: Recovering the Visual Space from Any Views](https://mlsanity.com/paper/2511.10647/) (2025)
- [Transformer Interpretability Beyond Attention Visualization](https://mlsanity.com/paper/2012.09838/) (2020)
- [LightGlue: Local Feature Matching at Light Speed](https://mlsanity.com/paper/2306.13643/) (2023)
- [Dissecting Recall of Factual Associations in Auto-Regressive Language Models](https://mlsanity.com/paper/2304.14767/) (2023)
- [Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass](https://mlsanity.com/paper/2501.13928/) (2025)
- [Towards Best Practices of Activation Patching in Language Models: Metrics and Methods](https://mlsanity.com/paper/2309.16042/) (2023)
- [A Tale of Two Features: Stable Diffusion Complements DINO for Zero-Shot Semantic Correspondence](https://mlsanity.com/paper/2305.15347/) (2023)
- [Understanding the Role of Individual Units in a Deep Neural Network](https://mlsanity.com/paper/2009.05041/) (2020)
- [Visual Geometry Grounded Deep Structure From Motion](https://mlsanity.com/paper/2312.04563/) (2023)
- [How to use and interpret activation patching](https://mlsanity.com/paper/2404.15255/) (2024)
- [What do Neural Machine Translation Models Learn about Morphology?](https://mlsanity.com/paper/1704.03471/) (2017)
- [Streaming 4D Visual Geometry Transformer](https://mlsanity.com/paper/2507.11539/) (2025)
- [SPair-71k: A Large-scale Benchmark for Semantic Correspondence](https://mlsanity.com/paper/1908.10543/) (2019)
- [VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold](https://mlsanity.com/paper/2505.12549/) (2025)
- [Emergent Outlier View Rejection in Visual Geometry Grounded Transformers](https://mlsanity.com/paper/2512.04012/) (2025)
- [Measuring the Interpretability of Unsupervised Representations via Quantized Reverse Probing](https://mlsanity.com/paper/2209.03268/) (2022)

## Built on by (most cited first, 7 of 7)

- [VGGT-SLAM 2.0: Real-time Dense Feed-forward Scene Reconstruction](https://mlsanity.com/paper/2601.19887/) (2026)
- [Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective](https://mlsanity.com/paper/2604.14025/) (2026)
- [SEAR: Simple and Efficient Adaptation of Visual Geometric Transformers for Unpaired RGB+Thermal 3D Reconstruction](https://mlsanity.com/paper/2603.18774/) (2026)
- [Self-Geometry: GT-Free and Plug-and-Play Test-Time Adaptation for Geometrically Consistent 3D Vision Foundation Models](https://mlsanity.com/paper/2608.10708/) (2026)
- [DDMS: Discriminative Distillation of Multi-view Foundational Features into Single-view Models](https://mlsanity.com/paper/2608.23850/) (2026)
- [GeoCond: A Conditioning-Aware Reliability Adapter for Feed-Forward 3D Reconstruction](https://mlsanity.com/paper/2609.18465/) (2026)
- [Argos: Adapt Rich Geometric Priors for Generalizable Online Scene-Change-Detection](https://mlsanity.com/paper/2610.10181/) (2026)

## Abstract (arXiv)

Feed-forward 3D reconstruction models such as DUSt3R, VGGT, and Depth Anything 3 (DA3) are transformer-based foundation models that infer camera geometry and dense scene structure in a single forward pass. Trained at scale in a supervised fashion, they raise a central question: do these models build upon geometric principles akin to traditional multi-view pipelines, or do they primarily rely on learned priors arising from the large-scale training setup? We find that epipolar geometry emerges within the intermediate layers of all three models and is causally linked to correspondence patterns in attention heads. To study this, we perform a systematic analysis of their internal representations across three real-world datasets and a controlled synthetic dataset. We quantify geometric understanding by probing intermediate features, analyzing attention patterns to identify correspondence matching patterns, and performing targeted interventions at the attention level. Further, we assess the role of learned priors by applying challenging input-level perturbations, such as occlusions, scene ambiguities, and varying camera configurations, and compare them against classical multi-stage reconstruction pipelines.