LIST3R: Long-sequence Instance-aware 3D Reconstruction

Jing Gao  ·  Wei Wang  ·  Feiran Wang  ·  Yan Yan

We present LIST3R, an instance-aware framework for long-sequence 3D reconstruction inspired by the way humans organize spatial memory around stable and recognizable objects. LIST3R organizes long-sequence reconstruction around instance anchors, using them to reconnect fragmented subsequences and consolidate local observations into a coherent global 3D scene. Given a long video, our approach partitions it into overlapping subsequences and builds a structured local instance library for each partial reconstruction, maintaining persistent trackable anchors with semantic and geometric evidence. These anchors are matched across subsequences to recover revisited regions and provide object-aware constraints for fragment alignment, producing a consistent global reconstruction. During this process, the evolving geometric evidence updates the local instance libraries and progressively organizes them into a unified global 3D instance library. Experiments on long-sequence benchmarks show that our method produces more accurate trajectories and higher-quality 3D reconstructions, highlighting the effectiveness of persistent instance anchors for organizing long-horizon 3D reconstruction.

LIST3R builds a local instance library for each subsequence, establishes cross-subsequence associations with instance-aware constraints, and consolidates local instance observations into a unified global 3D instance library.

Camera pose estimation on long sequences. ATE / RTE are reported in meters, RRE in degrees. Lower is better.

Point cloud reconstruction quality. Chamfer / Acc / Comp are reported in centimeters. Lower is better for Chamfer / Acc / Comp, and higher is better for NC / F@5.

@article{gao2026list3r,
  title={LIST3R: Long-sequence Instance-aware 3D Reconstruction},
  author={Gao, Jing and Wang, Wei and Wang, Feiran and Yan, Yan},
  journal={arXiv preprint arXiv:2607.00375},
  year={2026}
}