{ "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": "2609.25176", "title": "Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction", "summary": "The method replaces standard speech-to-speech assistant training with a three-stage pipeline: Core-Cocktail SFT plus multi-teacher on-policy distillation for reasoning, GRPO rollouts in self-evolving executable environments for tool use, and an alignment stage governing when to speak versus act. It costs large-scale RL infrastructure, executable tool environments, and multiple teacher models, none of which fit a single 24 GB GPU or a limited API budget. If adopted only via third-party API access to the resulting model, failure modes include dependence on vendor behavior, unverifiable safety and full-duplex claims on proprietary benchmarks, and no path to fine-tune coordination behavior for domain-specific tools.", "rationale": "This is a foundation-model training recipe requiring RL infrastructure and teachers the reader lacks, so it is only actionable if the vendor exposes the finished model through an API.", "technique": "Think-Act-Speak training for agentic realtime voice models", "family": "uso_de_ferramenta", "recommendation": "observar", "published": "2026-09-21", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.25176", "independent_implementations": 3, "total_implementations": 4, "stars": 288, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces standard speech-to-speech assistant training with a three-stage pipeline: Core-Cocktail SFT plus multi-teacher on-policy distillation for reasoning, GRPO rollouts in self-evolving executable environments for tool use, and an alignment stage governing when to speak versus act. It costs large-scale RL infrastructure, executable tool environments, and multiple teacher models, none of which fit a single 24 GB GPU or a limited API budget. If adopted only via third-party API access to the resulting model, failure modes include dependence on vendor behavior, unverifiable safety and full-duplex claims on proprietary benchmarks, and no path to fine-tune coordination behavior for domain-specific tools.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.25176", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Task success on the tau-Voice half-duplex adaptation rises from 78.4% to 82.0% versus Qwen-Audio-3.0-Realtime; false response rate to background speech on Full-Duplex-Bench v1.5 falls from 73.0% to 13.0%.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.25176", "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": "" }