SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation

Haoquan Fang, Markus Grotz, Wilbert Pumacay, Yi Ru Wang, Dieter Fox, Ranjay Krishna, Jiafei Duan
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:15925-15942, 2025.

Abstract

Robotic manipulation systems operating in diverse, dynamic environments must exhibit three critical abilities: multitask interaction, generalization to unseen scenarios, and spatial memory. While significant progress has been made in robotic manipulation, existing approaches often fall short in generalization to complex environmental variations and addressing memory-dependent tasks. To bridge this gap, we introduce SAM2Act, a multi-view robotic transformer-based policy that leverages multi-resolution upsampling with visual representations from large-scale foundation model. SAM2Act achieves a state-of-the-art average success rate of 86.8% across 18 tasks in the RLBench benchmark, and demonstrates robust generalization on The Colosseum benchmark, with only a 4.3% performance gap under diverse environmental perturbations. Building on this foundation, we propose SAM2Act+, a memory-based architecture inspired by SAM2, which incorporates a memory bank, an encoder, and an attention mechanism to enhance spatial memory. To address the need for evaluating memory-dependent tasks, we introduce MemoryBench, a novel benchmark designed to assess spatial memory and action recall in robotic manipulation. SAM2Act+ achieves an average success rate of 94.3% on memory-based tasks in MemoryBench, significantly outperforming existing approaches and pushing the boundaries of memory-based robotic systems. Project page: sam2act.github.io.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-fang25c, title = {{SAM}2{A}ct: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation}, author = {Fang, Haoquan and Grotz, Markus and Pumacay, Wilbert and Wang, Yi Ru and Fox, Dieter and Krishna, Ranjay and Duan, Jiafei}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {15925--15942}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/fang25c/fang25c.pdf}, url = {https://proceedings.mlr.press/v267/fang25c.html}, abstract = {Robotic manipulation systems operating in diverse, dynamic environments must exhibit three critical abilities: multitask interaction, generalization to unseen scenarios, and spatial memory. While significant progress has been made in robotic manipulation, existing approaches often fall short in generalization to complex environmental variations and addressing memory-dependent tasks. To bridge this gap, we introduce SAM2Act, a multi-view robotic transformer-based policy that leverages multi-resolution upsampling with visual representations from large-scale foundation model. SAM2Act achieves a state-of-the-art average success rate of 86.8% across 18 tasks in the RLBench benchmark, and demonstrates robust generalization on The Colosseum benchmark, with only a 4.3% performance gap under diverse environmental perturbations. Building on this foundation, we propose SAM2Act+, a memory-based architecture inspired by SAM2, which incorporates a memory bank, an encoder, and an attention mechanism to enhance spatial memory. To address the need for evaluating memory-dependent tasks, we introduce MemoryBench, a novel benchmark designed to assess spatial memory and action recall in robotic manipulation. SAM2Act+ achieves an average success rate of 94.3% on memory-based tasks in MemoryBench, significantly outperforming existing approaches and pushing the boundaries of memory-based robotic systems. Project page: sam2act.github.io.} }
Endnote
%0 Conference Paper %T SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation %A Haoquan Fang %A Markus Grotz %A Wilbert Pumacay %A Yi Ru Wang %A Dieter Fox %A Ranjay Krishna %A Jiafei Duan %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-fang25c %I PMLR %P 15925--15942 %U https://proceedings.mlr.press/v267/fang25c.html %V 267 %X Robotic manipulation systems operating in diverse, dynamic environments must exhibit three critical abilities: multitask interaction, generalization to unseen scenarios, and spatial memory. While significant progress has been made in robotic manipulation, existing approaches often fall short in generalization to complex environmental variations and addressing memory-dependent tasks. To bridge this gap, we introduce SAM2Act, a multi-view robotic transformer-based policy that leverages multi-resolution upsampling with visual representations from large-scale foundation model. SAM2Act achieves a state-of-the-art average success rate of 86.8% across 18 tasks in the RLBench benchmark, and demonstrates robust generalization on The Colosseum benchmark, with only a 4.3% performance gap under diverse environmental perturbations. Building on this foundation, we propose SAM2Act+, a memory-based architecture inspired by SAM2, which incorporates a memory bank, an encoder, and an attention mechanism to enhance spatial memory. To address the need for evaluating memory-dependent tasks, we introduce MemoryBench, a novel benchmark designed to assess spatial memory and action recall in robotic manipulation. SAM2Act+ achieves an average success rate of 94.3% on memory-based tasks in MemoryBench, significantly outperforming existing approaches and pushing the boundaries of memory-based robotic systems. Project page: sam2act.github.io.
APA
Fang, H., Grotz, M., Pumacay, W., Wang, Y.R., Fox, D., Krishna, R. & Duan, J.. (2025). SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:15925-15942 Available from https://proceedings.mlr.press/v267/fang25c.html.

Related Material