Catalogue reviewed 21 Aug 202637 evaluation assets across 18 financial-services workflows4 workflows without a direct public benchmarkLast ingest 20 Aug, 20:40 UK

OneBench evaluation position35 publicly reusable assets

A leaderboard is a screening tool, not a model approval.

Bank-grade evidence

Use public benchmarks to screen models; approve deployments with bank-owned use-case, control and operational tests.

Evaluation coverage matrix

Read across 18 workflows spanning customer, product, markets, finance, risk and enterprise operations. The taxonomy is condensed from the BIAN 14.0 service landscape ↗; every cell opens the catalogue at the relevant workflow and evaluation layer.

Public evidence is labelled direct, adjacent or absent

Evaluation asset coverage by financial-services workflow and evaluation-stack layer
Financial-services workflow	External screenPublic benchmarks and standards used to shortlist systems.	Golden setRepresentative institution-owned cases for the intended use.	ControlsSafety, conduct, security and policy-boundary testing.	OperationalWorkflow, tool-use, cost, latency and human-escalation evidence.	ContinuousRepeatable regression and production-change evaluation.
Customer & distribution
Customer service, complaints & conductIntent, service resolution, complaints, vulnerable customers and policy-bound conversations.Direct	3assets	Institution-owned	1assetAdjacent only	2assets	No mapped asset

Benchmarks and evals 101

Learn the concepts in four minutes: what a benchmark is, what an eval adds, how evidence moves from market comparison to bank approval, and how to challenge a headline score.

Open the full glossary →

The comparison layer

What is a benchmark?

A benchmark is a standardised test: the same tasks, conditions and scoring method are applied to multiple models or systems.

Useful for comparing options on a defined capability.
Strongest when the data, harness and scoring are inspectable.
Does not prove performance on your users, data or controls.

Open the glossary definition →

A bank-grade evaluation stack

01External screen

Use public benchmarks to shortlist models and expose obvious weaknesses.

Understand this layer →02Use-case golden set

Test representative, difficult and high-risk cases drawn from the bank's work.

Understand this layer →03Controls & adversarial

Probe leakage, entitlements, prompt injection, conduct, bias and unsafe actions.

Understand this layer →04Operational fitness

Measure latency, cost, availability, fallbacks, auditability and human escalation.

Understand this layer →05Continuous evaluation

Rerun on model, prompt, data, tool and policy changes; monitor production drift.

Understand this layer →

Benchmark and framework catalogue

Filter by financial-services workflow, capability area or evaluation type. Open any entry for measured capability, recommended banking use, metrics, limitation and primary source.

Primary sources linked

Search catalogueFinance workflow

Capability area

Evaluation type

Maturity

Access

Evaluation layer

5 matching evaluation assets0 selected for comparison


Public benchmark	Finance & investment, Legal & compliance, Operations & agents	Onboarding, KYC & customer due diligence, Lending, credit & collections +4	Growing	Open source	2026-08-21	The FinAI Open ↗
Public benchmark	Finance & investment, Legal & compliance, Operations & agents	Onboarding, KYC & customer due diligence, Payments, cards, fraud & scams +1	Growing	Open source	2026-08-21	Feedzai Research Open ↗
Public benchmark	Legal & compliance	Financial crime, sanctions & compliance, Internal audit & assurance +1	Established	Open dataset	2026-07-25	Stanford and legal contributors Open ↗
Public benchmark	Finance & investment, Knowledge & RAG, Operations & agents	Onboarding, KYC & customer due diligence, Corporate, commercial & trade finance +5	Emerging	Open source	2026-07-26	Databricks Open ↗
Assurance standard	Finance & investment, Legal & compliance, Operations & agents	Risk, model risk & stress testing, Financial crime, sanctions & compliance +1	Established	Public methodology	2026-08-21	Prudential Regulation Authority Open ↗

Public benchmark

CALM credit and risk benchmark

The FinAI

Growing

Open credit and risk evaluation covering credit scoring, fraud detection, financial distress and insurance claim classification across nine datasets.

Finance & investmentLegal & complianceOperations & agents

Relevant workflows

Onboarding, KYC & customer due diligence · Lending, credit & collections · Payments, cards, fraud & scams · Insurance underwriting, pricing & claims · Risk, model risk & stress testing · Financial crime, sanctions & compliance

Access: Open source
Last reviewed: 2026-08-21
Measures: 5 dimensions

What it measures

Credit risk classification
Fraud detection
Financial distress
Claim analysis
Bias

How a G-SIB should use it

Use as a tabular-risk and fairness screen before testing institution-specific lending, fraud, collections and underwriting decisions.

Useful metrics

Classification accuracyF1Class-balanced performanceGroup fairness

Primary sourceOpen CALM credit and risk benchmark

Evaluation and benchmark watch

Latest evidence from the existing OneBench source network that explicitly concerns model evaluation, benchmarks, assurance, red-teaming or measurable AI quality.

Updated with the main intelligence pipeline

arXiv cs.CL — Computation and Language11 Aug 2026

FinTrace: Holistic Trajectory-Level Evaluation of LLM Tool Calling for Long-Horizon Financial Tasks

FinTrace benchmark introduces trajectory-level evaluation for LLM tool-calling in long-horizon financial tasks, addressing limitations of call-level metrics.

Open source ↗

arXiv cs.CL — Computation and Language8 Aug 2026

Behavioral Canaries: Auditing Private Retrieved Context Usage in RL Fine-Tuning

Research proposes a new method, "Behavioral Canaries," to audit if private retrieved contexts are illicitly used in LLM RL fine-tuning.

Open source ↗

arXiv cs.CL — Computation and Language28 Apr 2026

FinGround: Detecting and Grounding Financial Hallucinations via Atomic Claim Verification

FinGround is a new research method to detect and ground financial hallucinations in LLMs by verifying atomic claims against regulatory filings, improving accuracy by 43%.

Open source ↗

arXiv cs.CL — Computation and Language28 Apr 2026

CiteAudit: You Cited It, But Did You Read It? A Benchmark for Verifying Scientific References in the LLM Era

Research identifies a benchmark, CiteAudit, to detect hallucinated citations from LLMs, which are present in scientific submissions.

Open source ↗

arXiv cs.LG — Machine Learning28 Apr 2026

Learning to Conceal Risk: Controllable Multi-turn Red Teaming for LLMs in the Financial Domain

Research introduces CoRT, a black-box multi-turn red-teaming framework to find concealed regulatory-violating risks in financial LLMs.

Open source ↗

arXiv cs.LG — Machine Learning

Methodology

OneBench includes a benchmark when its dataset, code or methodology is publicly inspectable, or when it provides a useful assurance pattern for a regulated enterprise. “Established” describes reuse and methodological maturity, not regulatory approval. Catalogue entries are editorial assessments; live stories link to the underlying publication. Public benchmark results should always be checked for dataset contamination, judge-model bias, version mismatch and unreported inference budgets.

Keep pace with finance AI evaluation

The daily briefing flags material benchmark and assurance developments. For catalogue governance or team access, speak to OneBench directly.

A leaderboard is a screening tool, not a model approval.

What is a benchmark?

CALM credit and risk benchmark

What it measures

How a G-SIB should use it

Useful metrics

FinTrace: Holistic Trajectory-Level Evaluation of LLM Tool Calling for Long-Horizon Financial Tasks

Behavioral Canaries: Auditing Private Retrieved Context Usage in RL Fine-Tuning

FinGround: Detecting and Grounding Financial Hallucinations via Atomic Claim Verification

CiteAudit: You Cited It, But Did You Read It? A Benchmark for Verifying Scientific References in the LLM Era

Learning to Conceal Risk: Controllable Multi-turn Red Teaming for LLMs in the Financial Domain

Keep pace with finance AI evaluation

The Price of Agreement: Measuring LLM Sycophancy in Agentic Financial Applications

One Size Fits None: Heuristic Collapse in LLM Investment Advice

When AI Speaks, Whose Values Does It Express? A Cross-Cultural Audit of Individualism-Collectivism Bias in Large Language Models

Rethinking XAI Evaluation: A Human-Centered Audit of Shapley Benchmarks in High-Stakes Settings

Algorithmic Compliance and Regulatory Loss in Digital Assets

When Graph Structure Becomes a Liability: A Critical Re-Evaluation of Graph Neural Networks for Bitcoin Fraud Detection under Temporal Distribution Shift

IndiaFinBench: An Evaluation Benchmark for Large Language Model Performance on Indian Financial Regulatory Text