aiqbench analysis
This research introduces OCC4M, a structured 4D memory system enabling vision-language models to perform robust spatiotemporal reasoning in long-horizon robotic manipulation tasks. By explicitly tracking objects across time and viewpoint changes rather than relying on raw observation history, the approach significantly outperforms baseline methods that struggle with state absence or camera shifts. The work advances the field by demonstrating how explicit memory structures can decouple high-level planning from low-level execution, improving success rates in complex environments.
OCC4M: Object-Centric 4D Memory for Spatiotemporal Reasoning in Long-Horizon Manipulation
arXiv:2609.28798v1 Announce Type: new Abstract: Long-horizon manipulation often requires reasoning about state absent from the current view, such as a vanished object's location, temporal identity, or the contents of a shuffled container. We present OCC4M ("Occam"), an object-centric 4D memory that maintains persistent tracks in a shared world frame and explicitly represents temporal, motion, and containment relations. A vision-language model (VLM) queries this structured memory to select actionable targets for history-free low-level execution. Across seven simulation conditions and 350 episodes, OCC4M achieves 96.6% memory success and 88.9% end-to-end success, versus 54.6% and 57.7% for FrameSamp, a raw-history VLM baseline using Gemini 3.7 Flash with the complete observation history and the same executor. In a controlled viewpoint-transfer test, OCC4M maintains 100% memory and 98% end-to-end success after a viewpoint change, while full-history FrameSamp falls to near-zero success. On 20 fixed-camera Franka episodes, OCC4M reaches 85% joint memory accuracy, versus at most 30% for FrameSamp across context sizes from $K=16$ to the complete history, and completes 45% of full two-stage tasks. These results support explicit object-centric memory for persistent spatiotemporal reasoning in long-horizon manipulation. Qualitative videos are available at https://occ4m-sup.github.io/occ4m-supplementary/.
Abstract:Long-horizon manipulation often requires reasoning about state absent from the current view, such as a vanished object's location, temporal identity, or the contents of a shuffled container. We present OCC4M ("Occam"), an object-centric 4D memory that maintains persistent tracks in a shared world frame and explicitly represents temporal, motion, and containment relations. A vision-language model (VLM) queries this structured memory to select actionable targets for history-free low-level execution. Across seven simulation conditions and 350 episodes, OCC4M achieves 96.6% memory success and 88.9% end-to-end success, versus 54.6% and 57.7% for FrameSamp, a raw-history VLM baseline using Gemini 3.7 Flash with the complete observation history and the same executor. In a controlled viewpoint-transfer test, OCC4M maintains 100% memory and 98% end-to-end success after a viewpoint change, while full-history FrameSamp falls to near-zero success. On 20 fixed-camera Franka episodes, OCC4M reaches 85% joint memory accuracy, versus at most 30% for FrameSamp across context sizes from $K=16$ to the complete history, and completes 45% of full two-stage tasks. These results support explicit object-centric memory for persistent spatiotemporal reasoning in long-horizon manipulation. Qualitative videos are available at this https URL.
13 pages, 11 figures, 6 tables. Supplementary videos: this https URL
Robotics (cs.RO)
arXiv:2609.28798 [cs.RO]
(or arXiv:2609.28798v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.28798 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 289 words
Related coverage
[AINews] not much happened today
Latent Space · Sep 10, 2026
VL-MemKnG: Hybrid Memory with a Spatio-Temporal Knowledge Graph for Question Answering over Long Egocentric Navigation Trajectories
arXiv · Jun 17, 2026
Humanity’s Last Invention — Richard Socher of Recursive
Latent Space · Sep 14, 2026
Expanding Spatial and Temporal Context for Robotic Imitation Learning With Scene Graphs
arXiv · Jun 8, 2026