{ "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": "2610.00388", "title": "T2SPO: Trajectory-to-Step Policy Optimization for Agentic Reinforcement Learning", "summary": "T2SPO replaces purely outcome-level reward in GRPO-style agentic RL with auxiliary per-step credit derived from a TabPFN regressor that estimates remaining distance to success from prior successful trajectories. The cost is maintaining a trajectory buffer and fitting/inferencing a TabPFN model at every step of every rollout, which adds pipeline complexity and inference overhead to an already expensive RL training loop. It can fail when successful trajectories are too sparse to give the regressor a useful context early in training, when the distance estimates are miscalibrated for out-of-distribution states, and the gains are validated only on ALFWorld and WebShop, so transfer to other agent environments is unverified.", "rationale": "The method requires an existing agentic RL training pipeline (GRPO on multi-billion-parameter models) that exceeds a single 24 GB GPU at the 7B scale, and the abstract provides no quantified improvement to justify adoption before independent results appear.", "technique": "Step-level credit assignment via TabPFN distance estimator", "family": "treino_eficiente", "recommendation": "observar", "published": "2026-09-30", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.00388", "independent_implementations": 0, "total_implementations": 1, "stars": 10, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "T2SPO replaces purely outcome-level reward in GRPO-style agentic RL with auxiliary per-step credit derived from a TabPFN regressor that estimates remaining distance to success from prior successful trajectories. The cost is maintaining a trajectory buffer and fitting/inferencing a TabPFN model at every step of every rollout, which adds pipeline complexity and inference overhead to an already expensive RL training loop. It can fail when successful trajectories are too sparse to give the regressor a useful context early in training, when the distance estimates are miscalibrated for out-of-distribution states, and the gains are validated only on ALFWorld and WebShop, so transfer to other agent environments is unverified.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.00388", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "\"T2SPO consistently improves overall task success over GRPO\" with 1.5B and 7B models on ALFWorld and WebShop; no quantitative figure is given in the abstract.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.00388", "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": "" }