MemBodied

Problem

To put a block back where it started, a robot has to remember where that was. Two episodes can reach the same camera image and still need different next actions. MemBodied asks how a vision-language-action policy can carry earlier evidence forward without storing an ever-growing history of frames.

Method

MemBodied combines recurrent associative memory with a fixed anchor of the initial scene. The policy queries the associative state, and a gated readout enters a dedicated memory token in the action expert. A separate pathway lets the current observation query the initial-scene anchor.

After the robot executes an action chunk, the next observation reveals its outcome. A delayed write associates the policy state with both the action and its observed consequence. Learned retention and write gates control a delta-rule update. The memory pathways are trained through the policy’s action objective; the episodic memory footprint stays constant in episode length for a fixed model configuration.

Simulation results

Across five RMBench tasks, with 50 rollouts per task and policy, the project reports 50.0% mean success for MemBodied and 6.4% for stateless π₀. Frame stacking reaches 14.8% and vanilla recurrent memory 16.8%.

The comparison with video-history memory includes both NativeMEM alone (38.4%) and NativeMEM integrated with MemBodied (45.2%). Running NativeMEM with and without MemBodied separates the effect of the video encoder from that of the recurrent memory. On LIBERO-Long, success rises from 85.2% to 90.6%.

Real-robot trials

On three physical robot tasks, each evaluated over 20 trials per policy, MemBodied succeeds in 16 of 60 trials (26.67%), compared with 2 of 60 (3.33%) for the stateless baseline. These tasks test remembering an original position, arrangement or order. The project page shows selected trials next to the totals.

MemBodied still fails in most real-robot trials. A fixed-size memory also forgets: the learned state keeps only what training taught it to keep.

The project page has the full method, results and robot videos. BaRe-Mem uses memory for a different job: remembering which sources to trust.