{ "schema_version": 1, "builder_plan": { "scenario": "You already need task-specific training and want to reduce its memory or compute cost.", "change": "Change only the training technique on a small controlled run after checking prerequisites.", "baseline": "Your current training recipe with the same model, data split, and evaluation.", "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": "Held-out quality, peak memory, training time, and full compute cost to the target quality.", "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": "2610.01509", "title": "Sharpening Tax in Post-Training", "summary": "PTGS replaces fixed-temperature sampling during RL post-training of agentic LLMs with a Bayesian sampler that adapts temperature per prompt to estimated difficulty, aiming to preserve pass@K solution coverage that standard RL post-training erodes. It costs additional estimation overhead per prompt and requires the infrastructure to run RL post-training in agentic environments, which exceeds a single 24 GB GPU and is not available through third-party inference APIs. It can fail if the difficulty estimate is noisy on few rollouts, and its benefit is demonstrated only in two agentic environments, so transfer to other task distributions is unverified.", "rationale": "The technique requires running RL post-training on agentic tasks, which the reader's constrained infrastructure cannot support, so its actionable takeaway is limited to preferring pass@K evaluation of base models over assuming post-trained checkpoints are strictly better.", "technique": "Posterior-tempered group sampling (PTGS) for RL post-training", "family": "treino_eficiente", "recommendation": "observar", "published": "2026-10-01", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.01509", "independent_implementations": 2, "total_implementations": 3, "stars": 136, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "PTGS replaces fixed-temperature sampling during RL post-training of agentic LLMs with a Bayesian sampler that adapts temperature per prompt to estimated difficulty, aiming to preserve pass@K solution coverage that standard RL post-training erodes. It costs additional estimation overhead per prompt and requires the infrastructure to run RL post-training in agentic environments, which exceeds a single 24 GB GPU and is not available through third-party inference APIs. It can fail if the difficulty estimate is noisy on few rollouts, and its benefit is demonstrated only in two agentic environments, so transfer to other task distributions is unverified.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.01509", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Across 14 base/post-trained model pairs on three agentic benchmarks, post-training systematically reduces pass@K solution coverage (the Sharpening Tax), and PTGS pays a smaller tax than fixed-temperature sampling, solving more tasks under repeated sampling while also improving pass@1; no multiplicative improvement factor is reported.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.01509", "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": "" }