Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
This paper introduces Latent Transition Memory (LTM), a novel in-context memory mechanism for vision-language-action models that encodes history as compact latent transitions between consecutive observations. LTM-$π$, a specific instantiation, leverages a frozen latent action encoder and causal KV caching for efficient, incremental inference without requiring semantic annotations or external reasoning modules. The method achieves state-of-the-art performance on benchmarks like RoboMME and LIBERO-Mem, demonstrating significant improvements in robotic manipulation tasks requiring long-term temporal reasoning and bounded-context operation.
Robotic agents can now retain crucial visual-motor history within a compact latent transition space, achieving near-perfect recall on complex manipulation tasks with minimal computational overhead.
Robotic manipulation often requires remembering interactions that are no longer visible. We introduce Latent Transition Memory (LTM), which stores history as compact latent transitions between consecutive observations. Instantiated on $π_{0.5}$ as LTM-$π$, our method uses a frozen latent action encoder, a lightweight adapter, and causal KV caching for incremental inference without memory-specific semantic annotations or external reasoning-model calls. With DreamDojo transitions, LTM-$π$ achieves 58.54% success on RoboMME and 80.5% on LIBERO-Mem, performing best among baselines without external reasoning models or memory-specific annotations. Ablations support the contribution of transition content beyond temporal cues and the efficiency of delayed memory admission. Scaling RoboMME demonstrations tenfold improves success from 57.67% to 87.50%, while a Cup Swap diagnostic highlights the role of process supervision in temporal composition. Under a 200-token memory budget, success remains 84.00%, versus 87.50% without compression. Finally, real-robot counting success improves from 11.25% to 82.50%. Together, these results demonstrate the effectiveness, scalability, and efficiency of latent transition memory, highlighting its capacity for complex temporal information, bounded-context operation, and real-world robotic control.