{ "schema_version": 1, "builder_plan": { "scenario": "An assistant chooses the wrong API or sends arguments your product cannot use.", "change": "Apply the proposed tool-selection or argument-generation change in a sandbox.", "baseline": "Your current agent with the same tool schemas and task budget.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Correct completed actions, invalid calls, retries, latency, and cost per completed task.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2610.31013", "title": "Multi-Tool Selection for Robotic Systems: Comparing Semantic Retrieval and LLM", "summary": "The method replaces direct large-model tool selection with clause-aware aggregated semantic retrieval over ChromaDB to choose the smallest complete subset of executable robot capabilities. It costs roughly 10 points of composite routing quality (0.7822 versus 0.8869) and requires maintaining an embedded tool registry and annotated routing data, in exchange for lower latency, local deployment, and no per-call API cost. It can fail on compound instructions through incomplete selection that omits necessary sensing or monitoring tools, or through over-selection that exposes irrelevant capabilities, especially as the tool registry grows beyond the 164-prompt evaluation regime.", "rationale": "The approach is deployable on constrained infrastructure and the quality-latency trade-off is explicitly quantified, but whether the 10-point quality gap is acceptable depends on the reader's specific tool set and prompt complexity distribution, so it warrants a workload-specific pilot rather than direct adoption.", "technique": "Clause-aware semantic retrieval for tool routing", "family": "uso_de_ferramenta", "recommendation": "testar", "published": "2026-10-22", "as_of": "2026-10-23", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.31013", "independent_implementations": 0, "total_implementations": 1, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces direct large-model tool selection with clause-aware aggregated semantic retrieval over ChromaDB to choose the smallest complete subset of executable robot capabilities. It costs roughly 10 points of composite routing quality (0.7822 versus 0.8869) and requires maintaining an embedded tool registry and annotated routing data, in exchange for lower latency, local deployment, and no per-call API cost. It can fail on compound instructions through incomplete selection that omits necessary sensing or monitoring tools, or through over-selection that exposes irrelevant capabilities, especially as the tool registry grows beyond the 164-prompt evaluation regime.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.31013", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "The cloud-hosted 120B LLM baseline achieves a completeness-weighted macro score of 0.8869 versus 0.7822 for the best retrieval-only method (CDB-v3), which in turn offers substantially lower latency and local deployment without cloud connectivity.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.31013", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_fetched", "core_kind": "", "equations_fetched_at": "", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }