AURA: Action-Gated Memory for Robot Policies at Constant VRAM
Quick Take
AURA-Mem introduces a constant-size recurrent memory for robots, optimizing memory writes by 5.19-9.19 times compared to traditional KV-caches. It maintains accuracy while using only 4,224 bytes of memory, outperforming ungated policies in efficiency without sacrificing performance on the LIBERO-Long benchmark.
Key Points
- AURA-Mem uses a learned gate to optimize memory writes based on action changes.
- It achieves 5.19-6.13 times fewer writes on synthetic benchmarks compared to O(1) baselines.
- The model maintains a fixed memory size of 4,224 bytes regardless of episode length.
- AURA-Mem matches ungated policies in success rate while reducing write frequency significantly.
- The method demonstrates a bound on approximate-information-state value-loss at scale.
Article Content
From source RSS / original summaryarXiv:2606. 02775v1 Announce Type: new Abstract: The KV-cache is the right memory for datacenters but the wrong memory for robots. Datacenter inference batches many short requests and resets them, amortizing an attention cache across a crowd. Embodied agents instead run one long, non-resetting episode on bandwidth-limited edge hardware, where high-bandwidth memory and flash are scarce, flash has finite write endurance, and memory writes rather than compute can become the binding constraint.
AURA-Mem (Action-Utility Recurrent Adaptive Memory) targets this regime. It wraps a frozen vision-language-action backbone with a constant-size recurrent memory and a learned gate that writes only when the current observation would change the next action: memory that knows when to stay silent. Unlike reconstruction-based memory, the gate is trained directly against a closed-loop action-error signal.
Its inference state is fixed at 4,224 bytes regardless of horizon, while a KV-cache grows to 6,061 times larger at 100,000 steps. On a controlled synthetic benchmark, AURA-Mem matches the best O(1) baseline in accuracy while using 5. 19-6. 13 times fewer writes, and up to 9. 19 times fewer writes on easier configurations. Budget-matched random and periodic schedules do not recover this gain, isolating the benefit to the action-surprise signal.
On a trained closed-loop OpenVLA-OFT 7B panel on LIBERO-Long (n=60 episodes per arm), the gate does not hurt success: AURA-Mem matches the ungated base policy (0. 233) and slightly exceeds an always-write KV arm (0. 217), while using 7. 0 times fewer writes and constant memory. We also instantiate an approximate-information-state value-loss bound as a methodology demonstration; at this scale, the bound is vacuous rather than a guarantee.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →AUDITFLOW: Executable Symbolic Environments for Structured Financial Reporting Verification
AuditFlow introduces a multi-agent framework for structured financial reporting verification, achieving 82.09% accuracy with GPT-5.5, outperforming the baseline by 14.93 points. It utilizes a symbolic environment for effective audit processes, demonstrating the necessity of deterministic checks for reliable verification.
