Benchmark RadarRusFinChain
RusFinChain is a Russian-language benchmark for verifiable chain-of-thought reasoning in finance, comprising 5,280 parameterized examples generated from executable Python templates across 17 domains and 172 topics. Each example includes a gold-standard reasoning chain with intermediate numeric values for automatic verification, and the dataset, code, and evaluation framework are publicly released.
- Released
- 2026-07-01
- Readiness
- Paper only
- Primary field
- Finance & Economics
Why it matters
Most financial reasoning benchmarks lack step-level supervision and are English-only; RusFinChain addresses this gap by providing a contamination-free, verifiable CoT benchmark for Russian, enabling automatic assessment of intermediate reasoning steps. It introduces improved metrics that better correlate with final-answer correctness, supporting more diagnostic evaluation of financial LLMs for the Russian-speaking community.
Motivation
Multi-step symbolic reasoning is essential for robust financial analysis, yet most benchmarks neglect intermediate reasoning steps.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.