Symbolic Mixture-of-Experts: Adaptive Skill-based Routing for Heterogeneous Reasoning
Justin Chih-Yao Chen, Sukwon Yun, Elias Stengel-Eskin, Tianlong Chen, Mohit Bansal
2025-03-11
Summary
This paper talks about Symbolic-MoE, a smart system that picks the best specialized AI models (like math or biology experts) for each question, then combines their answers to solve tough problems faster and better.
What's the problem?
Current AI systems either use one general model (which isn’t great at everything) or waste time loading many models for each task, making them slow and inefficient.
What's the solution?
Symbolic-MoE acts like a team manager: it quickly picks the right experts for each question (like algebra or biology specialists), groups similar questions together to load models only once, and combines their answers into one super-smart response.
Why it matters?
This makes AI tools better at solving complex problems (like medical diagnoses or math challenges) without needing tons of computers, helping experts and apps get faster, cheaper, and more accurate answers.
Abstract
Combining existing pre-trained expert LLMs is a promising avenue for scalably tackling large-scale and diverse tasks. However, selecting experts at the task level is often too coarse-grained, as heterogeneous tasks may require different expertise for each instance. To enable adaptive instance-level mixing of pre-trained LLM experts, we propose Symbolic-MoE, a symbolic, text-based, and gradient-free Mixture-of-Experts framework. Symbolic-MoE takes a fine-grained approach to selection by emphasizing skills, e.g., algebra in math or molecular biology in biomedical reasoning. We propose a skill-based recruiting strategy that dynamically selects the most relevant set of expert LLMs for diverse reasoning tasks based on their strengths. Each selected expert then generates its own reasoning, resulting in k outputs from k experts, which are then synthesized into a final high-quality response by an aggregator chosen based on its ability to integrate diverse reasoning outputs. We show that Symbolic-MoE's instance-level expert selection improves performance by a large margin but -- when implemented naively -- can introduce a high computational overhead due to the need for constant model loading and offloading. To address this, we implement a batch inference strategy that groups instances based on their assigned experts, loading each model only once. This allows us to integrate 16 expert models on 1 GPU with a time cost comparable to or better than prior multi-agent baselines using 4 GPUs. Through extensive evaluations on diverse benchmarks (MMLU-Pro, GPQA, AIME, and MedMCQA), we demonstrate that Symbolic-MoE outperforms strong LLMs like GPT4o-mini, as well as multi-agent approaches, with an absolute average improvement of 8.15% over the best multi-agent baseline. Moreover, Symbolic-MoE removes the need for expensive multi-round discussions, outperforming discussion baselines with less computation.