Fin-R1 — Open-Source Financial Reasoning LLM
China’s first open-source financial reasoning LLM (7B). FinEval score 75.2 (2nd place overall).
China’s first open-source financial reasoning LLM (7B). FinEval score 75.2 (2nd place overall).
Agent + RAG system for identifying high-net-worth scientists and generating personalized marketing strategies.
Production-grade unified moderation model: 92.3% accuracy, −35% false positives, +60% efficiency.
Open-source SOTA on MMEB-V2 (0.7523) and UVRB video retrieval (0.686).
Published in Under review at ACL 2026, 2026
A benchmark for evaluating the safety of financial agents in execution-grounded environments.
Recommended citation: Runguo Li*, et al. (2026). "FinVault: Benchmarking Financial Agent Safety in Execution-Grounded Environments." Under review at ACL 2026. (* First Co-Author)