cs.NI · 2606.18000 Copy arXiv ID · Jun 16, 2026 Save A T-API-Compliant ReAct Agentic Loop for Optical Networks: Generic vs. Domain-Specific Tool Abstractions Authors: Seyed Morteza Ahmadian , Paolo Monti , Carlos Natalino
Organizations: Department of Electrical Engineering, Chalmers University of Technology, Gothenburg, Sweden.
Abstract Optical networks need intent-driven, closed-loop agentic management, a key enabler for higher autonomy levels. We present the first T-API-compliant reasoning and act (ReAct) loop. We show that domain-specific composite tools achieve 90% oracle-validated correctness with threefold token savings compared to generic tools.
Explore similar work May 28, 2026 · Lorenz Kutschka, Bernhard Geiger Agentic Benchmarks
Feb 27, 2026 · Zhizhou He, Yang Luo, Xinkai Liu +4 Ai-Ran Self-Evolving Agents
May 28, 2026 · cs.AI J/K move · Enter open · S save
Lorenz Kutschka, Bernhard Geiger
Large language models in Agentic AI systems consume tool schemas and execution results and emit tool invocations as structured data. The default language for that exchange, JSON, was designed for application-to-application interchange rather than token efficiency, so its structural elements impose substantial token overhead. Recent work proposes token-optimized alternatives such as TOON (Token-Oriented Object Notation) and TRON (Token Reduced Object Notation) as more compact replacements, but these formats have been evaluated only on isolated comprehension or generation tasks. Whether their token reductions hold inside end-to-end agentic loops therefore remains an open question. We evaluate TOON and TRON on four agentic benchmarks (BFCL, MCPToolBenchPP, MCP-Universe, StableToolBench) and five open-weight LLMs, decoupling input compression from output compression to measure comprehension and generation independently. TRON reduces tokens by up to 27% with accuracy within 14pp of the JSON baseline. TOON achieves up to 18% reduction at a similar 9pp accuracy cost, but additionally cascades on multi-turn parsing failures and collapses parallel tool-call output for most models. The code is available at: https://github.com/lkutschka/notation-matters