Pyramid MoA: A Probabilistic Framework for Cost-Optimized Anytime Inference
This addresses the problem of high deployment costs for state-of-the-art LLMs, offering a cost-effective solution for high-volume applications, though it is incremental as it builds on existing ensemble and routing techniques.
The paper tackles the trade-off between inference cost and reasoning capability in Large Language Models by proposing Pyramid MoA, a hierarchical Mixture-of-Agents architecture that uses a lightweight Router to escalate queries only when necessary, achieving 93.0% accuracy on GSM8K while reducing compute costs by 61% compared to an Oracle baseline.
Large Language Models (LLMs) face a persistent trade-off between inference cost and reasoning capability. While "Oracle" models (e.g., Llama-3-70B) achieve state-of-the-art accuracy, they are prohibitively expensive for high-volume deployment. Smaller models (e.g., 8B parameters) are cost-effective but struggle with complex tasks. In this work, we propose "Pyramid MoA", a hierarchical Mixture-of-Agents architecture that uses a lightweight Router to dynamically escalate queries only when necessary. By leveraging semantic agreement and confidence calibration among an ensemble of small models, our Router identifies "hard" problems with high precision. On the GSM8K benchmark, our system achieves 93.0% accuracy, effectively matching the Oracle baseline (98.0%) while reducing compute costs by 61%. We demonstrate that the system introduces negligible latency overhead (+0.82s) and allows for a tunable trade-off between performance and budget.