{"slug": "finskillbench-evaluating-ai-agents-and-domain-skills-for-investment-management", "title": "FinSkillBench: Evaluating AI Agents and Domain Skills for Investment Management", "summary": "FinSkillBench, a new evaluation suite introduced by researchers on arXiv (2608.18099v1), measures whether language model agents can use financial domain skills for investment management tasks, spanning portfolio construction, risk management, and fundamental analysis across 12 subtasks and 2,603 task episodes. Testing 9 models, curated skill packages raised mean scores from 0.366 to 0.528, while self-generated skills provided little benefit despite higher computational cost. An independent evaluation with Hermes Agent (8 models, 5,280 episodes) reproduced the directional pattern, indicating that reliable procedural skills can be as important as model choice in investment management agents.", "body_md": "arXiv:2608.18099v1 Announce Type: new\nAbstract: Investment management is a high-stakes domain in which agentic AI systems must do more than generate plausible text. They must retrieve point-in-time data, assemble correct computational inputs, invoke specialized methods, and produce auditable structured outputs. We introduce FinSkillBench, an evaluation suite designed to measure whether language model agents can effectively use financial domain skills to solve investment management tasks. The benchmark spans three domains, portfolio construction, risk management, and fundamental analysis, and includes 12 subtasks with 2,603 task episodes.\nEach episode provides point-in-time inputs, hidden ground truth, and a task-specific verifier.We compare three conditions: no skill, curated skill packages consisting of procedural documents and executable components, and self-generated skills in which the agent writes and reuses its own procedures within an episode. Across 9 models and a large-scale evaluation, curated skills consistently improve performance, raising mean scores from 0.366 to 0.528, with the largest gains in portfolio construction and risk management.\nIn contrast, self-generated skills provide little benefit despite higher computational cost. An independent evaluation using a separate agent framework (Hermes Agent, 8 models, 5,280 episodes total) reproduces the directional pattern across all three domains, with the magnitude of skill effects varying by subtask and harness.\nThese results showthat in investment management agents, access to reliable procedural skills can be as important as model choice, while naive self-generation of skills is often ineffective. We release the benchmark, evaluation tools, curated skill packages, and full trajectories to support further research.", "url": "https://wpnews.pro/news/finskillbench-evaluating-ai-agents-and-domain-skills-for-investment-management", "canonical_source": "https://arxiv.org/abs/2608.18099", "published_at": "2026-08-20 04:00:00+00:00", "updated_at": "2026-08-20 04:15:28.669241+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-research", "large-language-models"], "entities": ["FinSkillBench", "arXiv", "Hermes Agent"], "alternates": {"html": "https://wpnews.pro/news/finskillbench-evaluating-ai-agents-and-domain-skills-for-investment-management", "markdown": "https://wpnews.pro/news/finskillbench-evaluating-ai-agents-and-domain-skills-for-investment-management.md", "text": "https://wpnews.pro/news/finskillbench-evaluating-ai-agents-and-domain-skills-for-investment-management.txt", "jsonld": "https://wpnews.pro/news/finskillbench-evaluating-ai-agents-and-domain-skills-for-investment-management.jsonld"}}