Akashic: A Low-Overhead LLM Inference Service with MemAttention
Quick Answer
Akashic introduces a low-overhead memory system utilizing MemAttention, enhancing LLM inference by organizing context into manageable chunks.
Quick Take
Akashic introduces a low-overhead memory system utilizing MemAttention, enhancing LLM inference by organizing context into manageable chunks. It achieves up to 10.2-point accuracy improvement, 1.21x throughput increase, and 1.88x sustainable request rate across various workloads compared to existing memory solutions.
Key Points
- Akashic organizes context into bounded chunks to improve efficiency.
- MemAttention models semantic relationships across these chunks.
- Achieves up to 10.2-point accuracy improvement in task performance.
- Increases throughput by up to 1.21x and request sustainability by 1.88x.
- Designed to reduce retrieval fragmentation and I/O overhead.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Recent LLM-based agent systems continuously accumulate context across multi-turn interactions, tool invocations, and cross-session workflows. Replaying the full history for every request quickly becomes impractical: long contexts increase prefill cost, may exceed context limits, and often bury task-relevant evidence in irrelevant content, degrading both serving efficiency and output quality. We propose Akashic, a low-overhead memory system built around MemAttention, which organizes context into bounded chunks and models semantic relationships across chunks, preserving cross-chunk evidence without repeatedly rewriting the full history. Akashic further applies hardware-software co-designed memory placement to co-locate likely co-retrieved chunks, reducing retrieval fragmentation and I/O overhead. Across four representative workloads and three model sizes, Akashic improves task accuracy by up to 10.2 points, throughput by up to 1.21x, and sustainable request rate by up to 1.88x over strong prior memory baselines.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.05708 [cs.AI] |
| (or arXiv:2607.05708v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.05708 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Junhao Hu [view email]
[v1]
Tue, 7 Jul 2026 00:06:22 UTC (2,337 KB)
— Originally published at arxiv.org
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 →Onnes: A Physics-Grounded LLM Simulator for Cryogenic Fault Diagnosis in Quantum Computing Infrastructure
Onnes is a physics-grounded digital twin simulator for dilution refrigerators, enhancing cryogenic fault diagnosis in quantum computing. It achieves a classification accuracy of 99.0% using few-shot demonstrations, matching a supervised ML classifier, while maintaining a low false alarm rate of 6.4% on real hardware.