{ "schema_version": 1, "builder_plan": { "scenario": "A self-hosted assistant is too expensive because its model uses most of the GPU memory.", "change": "Try the paper's supported lower-precision weights on a copy of your serving setup.", "baseline": "The same model at its original precision, on the same GPU.", "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": "Peak memory, task pass rate, latency, and total serving cost at the same context and load.", "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": "2610.08444", "title": "ActTune: Action-Aware Precision and GPU Operating-Point Adaptation for Energy-Efficient Vision-Language-Action Inference", "summary": "ActTune replaces fixed-precision (BF16) VLA inference with per-call, action-class-aware layer-wise precision allocation chosen by a decision tree, coupled with workload-dependent GPU frequency/power-cap selection, using a shared resident quantized weight bank to avoid weight reconstruction at switch time. It costs an offline calibration pass over configuration action errors, a latency-budgeted lookup-table build for GPU operating points, a forecasting controller running asynchronously, and up to 10% added inference latency. It can fail if the decision tree's action-error splits do not transfer to new tasks or action distributions, if workload forecasts mispredict and select poor operating points, or if the success and energy claims do not replicate outside LIBERO on different VLA models or hardware.", "rationale": "The method is reproducible on a single GPU and directly targets energy and latency costs of repeated VLA inference, but its gains are benchmark- and workload-specific, so it merits a validation trial on the reader's own robot tasks rather than immediate adoption or dismissal.", "technique": "Action-aware mixed precision with GPU operating-point tuning", "family": "quantizacao", "recommendation": "testar", "published": "2026-10-06", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.08444", "independent_implementations": 0, "total_implementations": 1, "stars": 2, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "ActTune replaces fixed-precision (BF16) VLA inference with per-call, action-class-aware layer-wise precision allocation chosen by a decision tree, coupled with workload-dependent GPU frequency/power-cap selection, using a shared resident quantized weight bank to avoid weight reconstruction at switch time. It costs an offline calibration pass over configuration action errors, a latency-budgeted lookup-table build for GPU operating points, a forecasting controller running asynchronously, and up to 10% added inference latency. It can fail if the decision tree's action-error splits do not transfer to new tasks or action distributions, if workload forecasts mispredict and select poor operating points, or if the success and energy claims do not replicate outside LIBERO on different VLA models or hardware.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.08444", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Up to 2.02x faster inference and up to 76.8% lower GPU energy per successful task versus BF16 baselines on LIBERO, with mean task success improved by up to 2.3% over state of the art and latency increase capped at 10%.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.08444", "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": "" }