{ "schema_version": 1, "builder_plan": { "scenario": "Your workload might fit a smaller or differently structured model.", "change": "Check for available weights and a compatible runtime before planning a product trial.", "baseline": "Your current model on identical held-out tasks and the same resource 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": "Task success, latency, memory, context limits, and migration cost.", "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": "2609.13045", "title": "Kraken: LLM-based Speech-to-Speech Translation via Low-bitrate VQ and Dual-path Source Conditioning", "summary": "Kraken replaces high-bitrate speech token prediction in speech LLMs with single-layer low-bitrate VQ tokens trained to reconstruct SSL features, plus a separate source-conditioned token-to-waveform decoder (Autowave-X) that handles non-linguistic transfer. It costs a full training pipeline: a Qwen3-8B backbone, 150k hours of multilingual multitask speech data, a custom VQ tokenizer, and a bespoke vocoder, which exceeds a 24 GB GPU and limited cloud budget. What can fail is reproduction at smaller data scale, loss of prosody or speaker fidelity from aggressive low-bitrate quantization, and dependence on a non-standard vocoder component that adds deployment complexity.", "rationale": "The technique depends on 150k hours of training data and a custom tokenizer-plus-vocoder stack built on an 8B LLM, so it is not adoptable under constrained infrastructure; the reader should instead consume such S2ST capability through third-party APIs if needed.", "technique": "Low-bitrate VQ speech tokens with dual-path source-conditioned vocoder", "family": "arquitetura_eficiente", "recommendation": "observar", "published": "2026-09-11", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.13045", "independent_implementations": 0, "total_implementations": 1, "stars": 1, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "Kraken replaces high-bitrate speech token prediction in speech LLMs with single-layer low-bitrate VQ tokens trained to reconstruct SSL features, plus a separate source-conditioned token-to-waveform decoder (Autowave-X) that handles non-linguistic transfer. It costs a full training pipeline: a Qwen3-8B backbone, 150k hours of multilingual multitask speech data, a custom VQ tokenizer, and a bespoke vocoder, which exceeds a 24 GB GPU and limited cloud budget. What can fail is reproduction at smaller data scale, loss of prosody or speaker fidelity from aggressive low-bitrate quantization, and dependence on a non-standard vocoder component that adds deployment complexity.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.13045", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "The paper reports better translation quality than SeamlessM4T-Large v2 and Qwen2.5-Omni, with improved speaker and prosody transfer, but provides no multiplicative factor in the abstract.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.13045", "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": [ { "anchor": "S3.E1", "label": "(1)", "section": "3 Method", "role": "loss", "latex": "\\mathcal{L}_{\\text{codec}}=\\underbrace{1-\\frac{\\bar{X}^{\\top}X}{\\|\\bar{X}\\|_{2}\\|X\\|_{2}}}_{\\text{cosine reconstruction loss}}+\\underbrace{\\left\\|z-\\mathrm{sg}(e_{k})\\right\\|_{2}^{2}}_{\\text{commitment loss}}.", "mathml": "ℒcodec=1−X¯⊤​X‖X¯‖2​‖X‖2⏟cosine reconstruction loss+‖z−sg⁡(ek)‖22⏟commitment loss.", "context": "", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" }, { "anchor": "S3.E2", "label": "(2)", "section": "3 Method", "role": "proposed_method", "latex": "\\mathcal{L}=p(\\bm{s}^{\\rm tgt}|\\bm{y}^{\\rm tgt},\\bm{y}^{\\rm src},X^{\\rm src};\\theta)\\\\\np(\\bm{y}^{\\rm tgt}|\\bm{y}^{\\rm src},X^{\\rm src};\\theta)p(\\bm{y}^{\\rm src}|X^{\\rm src};\\theta),", "mathml": "ℒ=p⁡(𝒔tgt|𝒚tgt,𝒚src,Xsrc;θ)p⁡(𝒚tgt|𝒚src,Xsrc;θ)​p​(𝒚src|Xsrc;θ),", "context": "We perform ASR and speech-to-text translation (S2TT) as intermediate steps to solve S2ST, instead of direct speech-to-speech prediction Rubenstein et al. (2023); Cheng et al. (2025); Futami et al. (2025). Let Xsrc denote the source language speech features, and 𝒔tgt denote the speech tokens of the target language. We consider predicting the transcribed text 𝒚src and 𝒚tgt as intermediate steps, formulated as:", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" } ], "equations_status": "selected", "core_kind": "formula", "equations_fetched_at": "2026-09-30", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }