{ "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": "2610.11655", "title": "Harness Evolution Hits a Ceiling: When Weight Training Should Begin", "summary": "It replaces undirected prompt engineering or defaulting to full fine-tuning for long-horizon agents with a diagnostic loop: label failed trajectories by first failure signal, evolve the harness to fix process failures, then train LoRA adapters on accepted trajectories to internalise the behaviour. Cost is a trajectory-labelling and harness-evolution loop plus single-GPU LoRA training on 4B-9B models, which fits constrained infrastructure but adds operational complexity in instrumentation and failure taxonomy. It can fail where gains live in runtime observations rather than model behaviour (WebArena-Lite adapters added nothing), the failure labelling misclassifies content versus process errors, or the evolved harness overfits to benchmark anchors rather than fresh tasks.", "rationale": "The method is feasible on a single 24 GB GPU (LoRA on 4B-9B models) and gives an actionable decision rule, but the gains are benchmark-specific and the transfer results show the recipe does not generalise uniformly, so workload-specific validation is required before production adoption.", "technique": "Failure-driven harness evolution with LoRA internalisation", "family": "treino_eficiente", "recommendation": "testar", "published": "2026-10-08", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.11655", "independent_implementations": 0, "total_implementations": 1, "stars": 122, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "It replaces undirected prompt engineering or defaulting to full fine-tuning for long-horizon agents with a diagnostic loop: label failed trajectories by first failure signal, evolve the harness to fix process failures, then train LoRA adapters on accepted trajectories to internalise the behaviour. Cost is a trajectory-labelling and harness-evolution loop plus single-GPU LoRA training on 4B-9B models, which fits constrained infrastructure but adds operational complexity in instrumentation and failure taxonomy. It can fail where gains live in runtime observations rather than model behaviour (WebArena-Lite adapters added nothing), the failure labelling misclassifies content versus process errors, or the evolved harness overfits to benchmark anchors rather than fresh tasks.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.11655", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Self-evolving harness lifts held-out DeepPlanning score of Qwen3.5-4B from 0.16 to 0.30 (9B from 0.32 to 0.44); LoRA adapters trained on evolved trajectories add +0.13 under the original harness, stack with the harness on 4B to more than double the held-out score, and match the full evolution line alone on 9B; +0.09 on 117 unseen WebArena-Lite tasks.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.11655", "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": "" }