{ "schema_version": 1, "builder_plan": { "scenario": "An agent looks good in demos, but you cannot tell which product tasks it fails.", "change": "Use the study's evaluation idea to inspect a small set of tasks with human-checked answers.", "baseline": "Your current evaluator and a manually reviewed reference set.", "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": "Agreement with human review, missed failures, false alarms, and evaluation cost per task.", "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": "2609.14992", "title": "MTAC-IFBench: Benchmarking Instruction-Following in Multi-Turn Agentic Coding", "summary": "MTAC-IFBench replaces ad hoc or single-turn evaluation of coding agents with a multi-turn benchmark averaging 7.04 turns and 91.33 constraints per instance, using per-item checklists verified by scripts and judge agents. Adopting it costs benchmark setup, API or GPU inference for the agents under test, and reliance on judge agents whose own reliability can distort scores. It can fail as an adoption decision if the reader's production workload differs from its 6 primary and 18 secondary constraint categories, since reported degradation over long sessions may not transfer.", "rationale": "The paper is an evaluation tool rather than a deployable capability, so its value to a constrained team is limited to validating their own coding agents on multi-turn instruction-following before production use.", "technique": "Multi-turn agentic coding instruction-following benchmark", "family": "avaliacao_de_agente", "recommendation": "testar", "published": "2026-09-14", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.14992", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "MTAC-IFBench replaces ad hoc or single-turn evaluation of coding agents with a multi-turn benchmark averaging 7.04 turns and 91.33 constraints per instance, using per-item checklists verified by scripts and judge agents. Adopting it costs benchmark setup, API or GPU inference for the agents under test, and reliance on judge agents whose own reliability can distort scores. It can fail as an adoption decision if the reader's production workload differs from its 6 primary and 18 secondary constraint categories, since reported degradation over long sessions may not transfer.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.14992", "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": "S4.Ex1", "label": "", "section": "4 Experiments", "role": "metric", "latex": "\\begin{gathered}\\begin{aligned} &\\mathrm{CSR}_{t}&&=\\frac{1}{|\\mathcal{C}_{t}|}\\sum_{j=1}^{|\\mathcal{C}_{t}|}z_{t,j}&\\qquad&\\mathrm{C\\text{-}ISR}_{t}&&=\\prod_{j=1}^{|\\mathcal{C}_{t}|}z_{t,j}\\\\[6.0pt]\n&\\mathrm{FSR}&&=\\frac{1}{|\\mathcal{F}|}\\sum_{k=1}^{|\\mathcal{F}|}y_{k}&\\qquad&\\mathrm{F\\text{-}ISR}&&=\\prod_{k=1}^{|\\mathcal{F}|}y_{k}\\end{aligned}\\\\[6.0pt]\n\\mathrm{BSR}=b\\end{gathered}", "mathml": "CSRt=1|𝒞t|​∑j=1|𝒞t|zt,jC​-​ISRt=∏j=1|𝒞t|zt,jFSR=1|ℱ|​∑k=1|ℱ|ykF​-​ISR=∏k=1|ℱ|ykBSR=b", "context": "ompletion of the final turn, let denote the executability of the generated project, denote the function checklist, and denote whether the -th function is satisfied. Following previous work (Qin et al. 2025; Qi et al. 2025; Ding et al. 2026), we define the following evaluation metrics. For instruction-following, CSRt (Constraint Success Rate) measures the proportion of satisfied constraints at turn , while C-ISRt (Constraint-Instruction Success Rate) measures strict completion where all constraints must be satisfied. For final functional correctness, BSR (Build Success Rate) measures whether the final generated project is executable, FSR (Function Success Rate) measures the proportion of satisfied functions, and F-ISR (Function-Instruction Success Rate) measures strict completion where all functional requirements must be satisfied. For each instance, these metrics are computed as follows:", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" } ], "equations_status": "selected", "core_kind": "formula", "equations_fetched_at": "2026-10-03", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }