{"slug": "dmaic-bench-from-self-reflection-to-self-restraint-for-agents-in-multi-stage", "title": "DMAIC-Bench: From Self-Reflection to Self-Restraint for Agents in Multi-Stage Analytical Tasks", "summary": "A new benchmark, DMAIC-Bench, finds that across 13 models from six providers, LLM agents average 92% routing accuracy on matched GO decisions but only 19% on NO-GO decisions, counting any forbidden downstream computation as a route error. The benchmark, built on Lean Six Sigma's statistical prerequisites, uses 97 seeded tasks across 15 industrial cases in four sectors plus 16 life-sciences tasks in RNA-seq and GWAS, and measures what the authors call self-restraint: whether evidence changes an agent's next action. Blinded expert review of 11 audited agents estimated that reports act on invalid evidence at 47% of reached NO-GO decisions, with an interval from 31% to 61%, and explicit stopping permission helped some models without consistently restoring correct decisions at both GO and NO-GO forks.", "body_md": "arXiv:2610.32305v1 Announce Type: cross \nAbstract: Multi-stage analytical tasks require LLM agents to use intermediate evidence to decide whether to continue, redirect or stop. Self-reflection, in which an agent critiques its output, is widely studied, yet a critique does not show whether evidence changed the next action. We call that capability self-restraint. Existing agent benchmarks largely measure task completion or output quality and leave it untested. DMAIC-Bench evaluates it with 97 seeded tasks across 15 industrial cases in four sectors, built around Lean Six Sigma's explicit statistical prerequisites. Paired counterfactual variants require opposite actions at the same workflow position, distinguishing warranted restraint from indiscriminate stopping. We assess execution traces with a private scorer and final claims through independent expert review. Across 13 models from six providers, execution-level routing accuracy averages 92% on matched GO decisions and 19% on NO-GO decisions, counting any forbidden downstream computation as a route error. Execution traces show agents reporting correct diagnostics before performing analyses that those results disallow. Separately, blinded expert review of the 11 audited agents estimates that reports act on invalid evidence at 47% of reached NO-GO decisions, with an interval from 31% to 61%. Explicit stopping permission helps some models without consistently restoring correct decisions at both GO and NO-GO forks. Sixteen life-sciences tasks in bulk ribonucleic acid sequencing (RNA-seq) and case-control genome-wide association studies (GWAS) show that the same framework can evaluate evidence-dependent decisions beyond industrial workflows. Project code and data are available at: https://github.com/psptpx/DMAIC-Bench-public.", "url": "https://wpnews.pro/news/dmaic-bench-from-self-reflection-to-self-restraint-for-agents-in-multi-stage", "canonical_source": "https://www.machinebrief.com/news/dmaic-bench-from-self-reflection-to-self-restraint-for-agent-evoj", "published_at": "2026-10-26 04:00:00+00:00", "updated_at": "2026-10-26 07:42:25.366814+00:00", "lang": "en", "topics": ["ai-agents", "ai-research", "large-language-models", "artificial-intelligence"], "entities": ["DMAIC-Bench", "Lean Six Sigma", "GitHub"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/dmaic-bench-from-self-reflection-to-self-restraint-for-agents-in-multi-stage", "markdown": "https://wpnews.pro/news/dmaic-bench-from-self-reflection-to-self-restraint-for-agents-in-multi-stage.md", "text": "https://wpnews.pro/news/dmaic-bench-from-self-reflection-to-self-restraint-for-agents-in-multi-stage.txt", "jsonld": "https://wpnews.pro/news/dmaic-bench-from-self-reflection-to-self-restraint-for-agents-in-multi-stage.jsonld"}}