SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach
英文标题：SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach
arXiv:2608.00030v1 Announce Type: new Abstract: Specialised retrieval agents typically surface higher quality results than general-purpose search, but selecting the optimal agent for a given query remains an open problem. Current approaches route queries based on inferred topic or intent, however i
2.6 KB
2.1 KB
640 B
612 B
380 B
9.1 KB
32.7 KB
22.8 KB
24.2 KB
34.7 KB
中文
arXiv:2608.00030v1 Announce Type: new Abstract: Specialised retrieval agents typically surface higher quality results than general-purpose search, but selecting the optimal agent for a given query remains an open problem. Current approaches route queries based on inferred topic or intent, however i
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
[Submitted on 15 Jul 2026]
Title:SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach
View a PDF of the paper titled SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach, by Gayathri V Kondapalli and 5 other authors
Abstract:Specialised retrieval agents typically surface higher quality results than general-purpose search, but selecting the optimal agent for a given query remains an open problem. Current approaches route queries based on inferred topic or intent, however intent-based selection is fundamentally limited: it does not incorporate signal from retrieved content, and cannot detect when a topically aligned agent produces low-relevance results. We address this by training a small language model via supervised fine-tuning followed by reinforcement learning to jointly perform agent selection and structured parameter generation for downstream tool calls, using a hierarchical reward function grounded in retrieval relevance along with query-agent topic alignment. This enables the model to learn task-dependent agent suitability from retrieval performance: which agents reliably yield high-relevance results for which query distributions, and when to redirect queries away from specialised agents despite surface-level topical overlap. On a targeted subset of such agent-query mismatches, the trained model achieves an NDCG@10 of 0.918 compared to 0.539 and 0.490 for two LLM baselines (Amazon Nova Lite and Claude Haiku 4.5) that route on intent alone. Overall, it achieves a mean NDCG@10 of 0.771 (+0.177 over Nova Lite, +0.219 over Haiku) with a mean selection latency of 120.1ms, an 82.4% reduction over Nova Lite.
Submission history
From: Venkatashesha Gayathri Kondapalli [
]
[v1]
Wed, 15 Jul 2026 11:11:39 UTC (129 KB)