RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies

Memory is critical for long-horizon and history-dependent robotic manipulation. Such tasks often involve counting repeated actions or manipulating objects that become temporarily occluded. Recent vision-language-action (VLA) models have begun to incorporate memory mechanisms; however, their evaluations remain confined to narrow, non-standardized settings. This limits systematic understanding, comparison, and progress measurement. To address these challenges, we introduce **RoboMME**: a large-scale standardized benchmark for evaluating and advancing VLA models in long-horizon, history-dependent scenarios. Our benchmark comprises 16 manipulation tasks constructed under a carefully designed taxonomy that evaluates *temporal*, *spatial*, *object*, and *procedural* memory. We further develop a suite of 14 memory-augmented VLA variants built on the $\pi_{0.5}$ backbone to systematically explore different memory representations across multiple integration strategies. We show that the effectiveness of memory representations is highly task-dependent, with each design offering distinct advantages and limitations across different tasks. Videos and code can be found at https://robomme.github.io

Robots often struggle with tasks that require memory, such as remembering where an object was placed, how many times something happened, or what action a person demonstrated earlier. We introduce RoboMME, a benchmark for evaluating whether robot learning models can remember, reason, and act over time. RoboMME includes manipulation tasks covering different forms of memory, including object locations, event counting, object references, and imitation from demonstrations. We use it to study several ways of adding memory to vision-language-action robot models, such as text summaries, stored visual information, and recurrent memory states. Our results show that memory improves robot performance, but different memory designs work best for different tasks. RoboMME provides a controlled testbed for developing robots that can handle longer and more realistic interactions.