benchmarkAI Agent Evaluation10d ago
DeepFinance Benchmark Released for Long-Horizon AI Agent Financial Tasks
Facts
DeepFinance is a new benchmark designed to evaluate AI models on long-horizon, multi-step private equity and investment banking tasks. It tests agent performance across 21 end-to-end tasks, measuring capabilities in judgment, workflow accuracy, and auditability against expert-written rubrics.
- ▸Benchmark includes 21 end-to-end tasks requiring an average of 27.1 hours for a human professional.
- ▸Data includes 19 files per task and uses 84 expert-written rubric criteria.
- ▸Evaluation covers 22 model configurations using native agent harnesses.
- ▸GPT-6 Astra with max reasoning achieved the highest performance at 57.6% adherence to criteria.
- ▸Identified systematic model failures in acting on decision-relevant evidence despite accurate calculations.
- Occurred
- 2026-11-07 19:31 UTC
- First reported
- 2026-11-07 19:31 UTC
- Last updated
- 2026-11-10 03:05 UTC
- Event type
- benchmark
- Region
- Global
Entities
Categories
Source timeline · 1
- arXivFirst-party2026-11-07 19:31
Community vote
Sign in to vote on importance and sentiment.
Platform Signals
Impact score
66
AI sentiment
NeutralEvent significance30%1
Entity importance20%1
Source credibility20%1
Novelty15%1
Corroboration15%1
Confidence1
Novelty1
Corroboration1
Community
ImportanceNo votes
Strategic sentiment
Positive 0%Neutral 0%Negative 0%
Platform signals are algorithmic and community-derived indicators. They are not investment recommendations and do not predict valuation or performance.
Related events
- 1dIntroduction of SciResearch-Eval Benchmark for AI Agent Scientific Reasoning
- 1dRelease of CAE-Bench to Evaluate AI Agents in Computer-Aided Engineering Solver Development
- 7dIntroduction of ExpertCrowd-SWE Benchmark for Evaluating Autonomous Coding Agents
- 8dSciMARS Framework Improves Multi-Agent Scientific Discovery
- 8dResearchers develop zero-search reinforcement learning agent for Slay the Spire 2