CPU-ARC: Agentic Reasoning Caching for on-CPU Reasoning Reuse
arXiv cs.AI · 7 Nov 2026, 04:58 AEST
A new method called agentic reasoning caching (ARC) allows for efficient reuse of reasoning from large language models. This approach demonstrates significant cost savings and performance comparable to leading models in complex tasks.
Key facts
- Large language models reinvent the wheel all the time.
- We show that once a task has been reasoned through, the application of its solution to a new problem instance can be done very efficiently and with little involvement from language models.
- ARCs show on-par performance with frontier LLMs across a range of difficult mathematical and scientific tasks, are fully explainable, continually trainable, robust to input perturbations, and 16-291x cheaper to invoke than the source model.
Read aloud
A new method called agentic reasoning caching ARC allows for efficient reuse of reasoning from large language models. This approach demonstrates significant cost savings and performance comparable to leading models in complex tasks. Large language models reinvent the wheel all the time. We show that once a task has been reasoned through, the application of its solution to a new problem instance can be done very efficiently and with little involvement from language models. ARCs show on-par performance with frontier L L M s across a range of difficult mathematical and scientific tasks, are fully explainable, continually trainable, robust to input perturbations, and sixteen-291x cheaper to invoke than the source model.
Sources in this cluster
- CPU-ARC: Agentic Reasoning Caching for on-CPU Reasoning Reuse
arXiv cs.AI · 7 Nov 2026, 04:58 AEST