{"slug": "can-you-trust-your-finance-agent-drift-diagnosis-and-auditable-intervention-on", "title": "Can You Trust Your Finance Agent? Drift, Diagnosis and Auditable Intervention on One Shipped Financial Research Agent", "summary": "A case study on the Vals Finance Agent Benchmark v2 reference agent found that seven identical runs on 18 questions passed between 105 and 123 of 149 rubric criteria, with cost varying by up to 17%, according to the arXiv paper 2611.17939v1. Classifying 20 failed trajectories from three runs of nine held-out questions, the authors attributed 13 failures to the agent's instructions, 3 to its tools, 4 to the model and none to its data, and four blind readers from three model families named the same fault as the LLM analyzer in 94% of cases, matching their 93% agreement with each other. An optimization loop that made one targeted change per round produced a final configuration costing 0.34x the baseline on held-out questions (95% CI [0.22, 0.58]) at accuracy within noise under a cost objective, while an accuracy objective gained 14 points on development questions but only 1.5 on held-out ones (CI [-5.3, +7.2]) — a gain the authors judged not real against measured run-to-run noise.", "body_md": "arXiv:2611.17939v1 Announce Type: new \nAbstract: Can you trust a finance agent you did not build? We take the reference agent shipped with the Vals Finance Agent Benchmark v2, leave it untuned, and ask four questions a leaderboard score does not answer. Is the score stable? Seven identical runs on 18 questions pass between 105 and 123 of 149 rubric criteria, and the bill varies by up to 17%. When the agent fails, who owns the fix? We classify 20 failed trajectories, from three runs of nine held-out questions, by the layer whose owner would remedy them, under a taxonomy frozen in advance: 13 belong to the agent's instructions, 3 to its tools, 4 to the model, none to its data. Can that diagnosis be trusted? Four blind readers from three model families, given only the traces, name the same fault as our LLM analyzer in 94% of cases, the rate at which they agree with one another (93%). Can a change be trusted? An optimization loop reads the traces, makes one targeted change per round and keeps it only if it clears a fixed threshold. Under a cost objective its final configuration costs 0.34x the baseline on the held-out questions (95% CI [0.22, 0.58]) at accuracy within noise. Under an accuracy objective it gains 14 points on the development questions and 1.5 on the held-out ones (CI [-5.3, +7.2]). Judged against the measured run-to-run noise, under a rule written before the test was opened, the first is a real saving and the second is not. This is a case study of one agent and nine test questions, not a general result.", "url": "https://wpnews.pro/news/can-you-trust-your-finance-agent-drift-diagnosis-and-auditable-intervention-on", "canonical_source": "https://www.machinebrief.com/news/can-you-trust-your-finance-agent-drift-diagnosis-and-auditab-pv8j", "published_at": "2026-11-16 05:00:00+00:00", "updated_at": "2026-11-16 05:25:30.420417+00:00", "lang": "en", "topics": ["ai-agents", "ai-safety", "ai-research", "large-language-models"], "entities": ["Vals Finance Agent Benchmark v2", "arXiv", "Vals"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/can-you-trust-your-finance-agent-drift-diagnosis-and-auditable-intervention-on", "markdown": "https://wpnews.pro/news/can-you-trust-your-finance-agent-drift-diagnosis-and-auditable-intervention-on.md", "text": "https://wpnews.pro/news/can-you-trust-your-finance-agent-drift-diagnosis-and-auditable-intervention-on.txt", "jsonld": "https://wpnews.pro/news/can-you-trust-your-finance-agent-drift-diagnosis-and-auditable-intervention-on.jsonld"}}