Paper 2604.14144
SpatialEvo: Self-Evolving Spatial Intelligence via Deterministic Geometric Environments
- Published
- Apr 2026
- Research lab
- Independent
- Citations
- 3
- GitHub
- 80 stars
01 In brief
Summary
SpatialEvo introduces a self-evolving framework for 3D spatial reasoning that replaces model consensus with deterministic geometric feedback.
The core contribution is the Deterministic Geometric Environment (DGE), which defines validation rules for 16 spatial reasoning task categories and computes exact ground truth from point clouds and camera poses, transforming unannotated 3D scenes into zero-noise reward judges.
A single policy model co-evolves as Questioner (generating valid questions) and Solver (deriving answers) under DGE constraints, trained via GRPO.
A task-adaptive scheduler dynamically adjusts sampling based on historical accuracy, creating an emergent curriculum.
Experiments across nine benchmarks show SpatialEvo achieves the highest average scores at both 3B and 7B scales (51.1 and 54.7), with consistent gains on spatial reasoning benchmarks and no degradation on general visual understanding.
Ablations confirm that replacing DGE ground truth with majority-vote pseudo-labels causes the largest performance drop, validating the importance of deterministic physical feedback.
02 From the paper
Abstract
Spatial reasoning over three-dimensional scenes is a core capability for embodied intelligence, yet continuous model improvement remains bottlenecked by the cost of geometric annotation. The self-evolving paradigm offers a promising path, but its reliance on model consensus to construct pseudo-labels causes training to reinforce rather than correct the model's own geometric errors. We identify a property unique to 3D spatial reasoning that circumvents this limitation: ground truth is a deterministic consequence of the underlying geometry, computable exactly from point clouds and camera poses without any model involvement. Building on this insight, we present SpatialEvo, a self-evolving framework for 3D spatial reasoning, centered on the Deterministic Geometric Environment (DGE). The DGE formalizes 16 spatial reasoning task categories under explicit geometric validation rules and converts unannotated 3D scenes into zero-noise interactive oracles, replacing model consensus with objective physical feedback. A single shared-parameter policy co-evolves across questioner and solver roles under DGE constraints: the questioner generates physically valid spatial questions grounded in scene observations, while the solver derives precise answers against DGE-verified ground truth. A task-adaptive scheduler endogenously concentrates training on the model's weakest categories, producing a dynamic curriculum without manual design. Experiments across nine benchmarks demonstrate that SpatialEvo achieves the highest average score at both 3B and 7B scales, with consistent gains on spatial reasoning benchmarks and no degradation on general visual understanding.