aiqbench analysis
This research introduces CM-LRS, a novel benchmark specifically designed for capital markets that evaluates LLM outputs on seven critical dimensions like evidence traceability and auditability rather than just surface-level accuracy. By testing models against real SEC filings across complex workflows such as M&A reasoning, the study demonstrates that frontier closed-source models significantly outperform open-weights baselines in generating "bankable" drafts defensible by regulators or counterparties. The findings highlight a persistent gap between plausibility and reliability in high-stakes financial applications where retrieval and synthesis capabilities are paramount.
Capital Markets LLM Reliability Score (CM-LRS): From Plausible to Bankable
arXiv:2607.21340v1 Announce Type: new Abstract: In capital-markets workflows the question is rarely whether a large language model can produce a fluent draft, but whether the draft is bankable: defensible in front of a counter-party or a regulator, with the documents in hand. Existing methods address parts of that gap: open-domain QA benchmarks reward surface accuracy, and finance benchmarks (FinanceBench, FinQA, ConvFinQA) advance document-grounded and numerical QA but evaluate at the question-answer layer rather than the workflow outputs practitioners defend. We introduce CM-LRS, a Capital Markets LLM Reliability Score, evaluating outputs at the workflow-output layer across seven dimensions: factual accuracy, evidence traceability, numerical consistency, workflow completeness, source discipline, decision usefulness, and reviewability/auditability. Each is scored 0-5 against a rubric anchored on signals reviewers in regulated settings use; the aggregate is tunable to the workflow. We demonstrate CM-LRS on five workflows (DCM transaction-terms extraction, precedent retrieval, issuer profile synthesis, M&A transaction-comparable reasoning, ECM transaction-terms extraction) over public SEC EDGAR filings, a public UK takeover release, and fictional synthetic supplements, scoring four models against four independent LLM judges spanning three model families. Three findings. First, the frontier closed-source models cluster within 0.22 points on four-judge averaged CM-LRS (Sonnet 4.6 = 4.31, Opus 4.7 = 4.30, GPT-5.5 = 4.09); all four judges place the open-weights baseline (Llama 3.3 70B = 3.15) last. Second, that gap concentrates on retrieval (2.23) and synthesis (2.15), not extraction (0.84). Third, Decision Usefulness shows the widest cross-model dispersion of any dimension (4.0 points on issuer profiling) and top-tier inter-judge agreement (mean r = 0.52). Plausibility is cheap. Bankability is the bar.
Abstract:In capital-markets workflows the question is rarely whether a large language model can produce a fluent draft, but whether the draft is bankable: defensible in front of a counter-party or a regulator, with the documents in hand. Existing methods address parts of that gap: open-domain QA benchmarks reward surface accuracy, and finance benchmarks (FinanceBench, FinQA, ConvFinQA) advance document-grounded and numerical QA but evaluate at the question-answer layer rather than the workflow outputs practitioners defend. We introduce CM-LRS, a Capital Markets LLM Reliability Score, evaluating outputs at the workflow-output layer across seven dimensions: factual accuracy, evidence traceability, numerical consistency, workflow completeness, source discipline, decision usefulness, and reviewability/auditability. Each is scored 0-5 against a rubric anchored on signals reviewers in regulated settings use; the aggregate is tunable to the workflow. We demonstrate CM-LRS on five workflows (DCM transaction-terms extraction, precedent retrieval, issuer profile synthesis, M&A transaction-comparable reasoning, ECM transaction-terms extraction) over public SEC EDGAR filings, a public UK takeover release, and fictional synthetic supplements, scoring four models against four independent LLM judges spanning three model families. Three findings. First, the frontier closed-source models cluster within 0.22 points on four-judge averaged CM-LRS (Sonnet 4.6 = 4.31, Opus 4.7 = 4.30, GPT-5.5 = 4.09); all four judges place the open-weights baseline (Llama 3.3 70B = 3.15) last. Second, that gap concentrates on retrieval (2.23) and synthesis (2.15), not extraction (0.84). Third, Decision Usefulness shows the widest cross-model dispersion of any dimension (4.0 points on issuer profiling) and top-tier inter-judge agreement (mean r = 0.52). Plausibility is cheap. Bankability is the bar.
23 pages. Resubmission of submit/7557765, which expired due to an arXiv system bug (confirmed in support ticket AH-199019); the overfull-box correction requested by moderators has been applied. Original submission held in moderation since 14 May 2026. Therefore, request priority review/approval
Computation and Language (cs.CL)
arXiv:2607.21340 [cs.CL]
(or arXiv:2607.21340v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2607.21340 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 316 words
Related coverage
The world's best gradual disempowerment model organism: Frontier AI labs
LessWrong · Sep 29, 2026
Debating RSI, the US-China Gap, and Jaggedness with JS Denain of Epoch AI
Interconnects · Sep 22, 2026
The death of humanity
Fortune · Oct 16, 2026
Building and evaluating model diffing agents
AI Alignment Forum · Jun 12, 2026