{"slug": "slms-as-multi-agent-routers-a-progressive-sft-and-reinforcement-learning", "title": "SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach", "summary": "Researchers trained a small language model to route queries to specialized retrieval agents, achieving an NDCG@10 of 0.918 on mismatched queries, compared to 0.539 for Amazon Nova Lite and 0.490 for Claude Haiku 4.5, and an overall mean NDCG@10 of 0.771 with a mean selection latency of 120.1ms, an 82.4% reduction over Nova Lite. The approach uses supervised fine-tuning followed by reinforcement learning to jointly perform agent selection and parameter generation, incorporating retrieval relevance and query-agent topic alignment.", "body_md": "arXiv:2608.00030v1 Announce Type: new\nAbstract: 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.", "url": "https://wpnews.pro/news/slms-as-multi-agent-routers-a-progressive-sft-and-reinforcement-learning", "canonical_source": "https://arxiv.org/abs/2608.00030", "published_at": "2026-08-04 04:00:00+00:00", "updated_at": "2026-08-04 04:35:56.947701+00:00", "lang": "en", "topics": ["machine-learning", "artificial-intelligence", "ai-agents", "natural-language-processing"], "entities": ["Amazon Nova Lite", "Claude Haiku 4.5"], "alternates": {"html": "https://wpnews.pro/news/slms-as-multi-agent-routers-a-progressive-sft-and-reinforcement-learning", "markdown": "https://wpnews.pro/news/slms-as-multi-agent-routers-a-progressive-sft-and-reinforcement-learning.md", "text": "https://wpnews.pro/news/slms-as-multi-agent-routers-a-progressive-sft-and-reinforcement-learning.txt", "jsonld": "https://wpnews.pro/news/slms-as-multi-agent-routers-a-progressive-sft-and-reinforcement-learning.jsonld"}}