Search papers, labs, and topics across Lattice.
This paper introduces Routed Graph Handoff, a novel approach that utilizes a lightweight LLM router to dynamically select between natural language and structured graphs for multi-agent delegation, significantly optimizing token usage. The method achieves substantial performance improvements across four benchmarks, demonstrating up to a 12.7 percentage point increase in task success while achieving compression ratios of up to 3.2 times. Notably, the router mitigates the performance drop seen with graph-only delegation, maintaining high efficacy at minimal overhead.
Switching between natural language and structured graphs can boost multi-agent LLM performance by over 12 percentage points while slashing token usage by more than threefold.
Multi-agent LLM systems coordinate through natural-language messages that consume 40--60\% of their token budget. Replacing these with structured graphs reduces cost but fails on tasks requiring adaptive reasoning. We propose \textbf{Routed Graph Handoff}, where a lightweight LLM router (155 tokens, 0.15\% overhead) selects between a typed dependency graph and natural language for each delegation. On four benchmarks (1,050+ trajectories), the routed system matches or exceeds NL-only on every task: \textbf{+12.7\,pp} on $\tau$-retail at 3.2$\times$ compression ($p{<}0.01$), \textbf{+8.7\,pp} on BrowseComp at 2.2$\times$ compression ($p{<}0.05$), and parity on BFCL and AppWorld. Without the router, graph-only delegation regresses 14.6\,pp on AppWorld; the router eliminates this at near-zero cost. A graph-aware executor prompt is required: the same schema without interpretation guidance yields no gain. An oracle analysis reveals 8.6\,pp of additional headroom, motivating execution-time adaptive routing as future work.
