Agent-conditioned routing doubles diverse solution paths in AI models
Agent-Conditioned Routing: Diverse Reasoning Paths Within One Frozen Model
Artificial IntelligenceMachine LearningMultiagent Systems
Summary
Generating multiple ways to solve math and physics problems often results in similar answers. The authors studied a frozen AI model that can create diverse solutions without changing the model or using special prompts. They introduced Agent-Conditioned Routing (ACR), which assigns different fixed masks to create varied reasoning paths in parallel. Tests show ACR produces almost twice as many distinct solution methods compared to traditional techniques, with minimal extra computation.
What this means in practice
- •For ai engineers: Generate multiple distinct solution methods from a single AI model to improve problem-solving diversity without retraining or copying models.
- •For machine learning platform operators: Integrate agent-conditioned routing to multiply reasoning paths with minimal extra resource use and latency impact during batched inference.
Authors
Ananth Kalyanasundaram, Ankush Kadu, Athira Gopal, Aswanth Krishnan
Abstract
Generating several solutions can expose alternative methods, but additional trajectories often paraphrase one dominant approach. We investigate whether the conditional capacity of a frozen sparse mixture-of-experts (MoE) model can provide such diversity from one problem prompt, without user-assigned roles, prescribed methods, fine-tuning, or model copies. Agent-Conditioned Routing (ACR) assigns $m$ parallel trajectories persistent, untrained shared/private expert-eligibility masks while preserving the learned, input-dependent token router. On a screened 126-question multi-mode mathematics and physics benchmark, with model judges reading the reported derivations in full, we count the number of problems yielding at least two distinct primary methods among soft-correct solutions in four trajectories. ACR reaches approximately twice as many such problems as the strongest temperature setting under GLM 5.3 (37.3 versus 18.7), and 1.67 times as many under GPT-5.6 Sol (45.00 versus 27.00), at matched output budgets. A blinded four-expert audit on a stratified sample of 30 problems also finds more multi-method solutions with ACR on both backbones, with lower correctness. Matched-eligibility ablations show modest multi-method count advantages for ACR, while token-resampled masks attain higher correctness. An informed sequential prompt elicits more diversity still, using previous solutions and explicit requests for alternative methods. ACR adds about 160 KiB of mask state. In a three-problem microbenchmark capped at 1,024 tokens, its latency and peak VRAM usage remain within 3% of batched temperature sampling.