# DFAH-Bench: Benchmarking Observable Agent Instability in Financial Decision-Making

arXiv: https://arxiv.org/abs/2607.20491 · HTML: https://mlsanity.com/paper/2607.20491/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 39 references that are in the map.
- First posted on arXiv 10 Jun 2026.

## Builds on (most cited first, 20 of 20)

- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models](https://mlsanity.com/paper/2206.04615/) (2022)
- [Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift](https://mlsanity.com/paper/1906.02530/) (2019)
- [BloombergGPT: A Large Language Model for Finance](https://mlsanity.com/paper/2303.17564/) (2023)
- [Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?](https://mlsanity.com/paper/2004.03685/) (2020)
- [ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilities](https://mlsanity.com/paper/2408.04682/) (2024)
- [FinBen: A Holistic Financial Benchmark for Large Language Models](https://mlsanity.com/paper/2402.12659/) (2024)
- [Visibility into AI Agents](https://mlsanity.com/paper/2401.13138/) (2024)
- [TRAJECT-Bench:A Trajectory-Aware Benchmark for Evaluating Agentic Tool Use](https://mlsanity.com/paper/2510.04550/) (2025)
- [τ-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge](https://mlsanity.com/paper/2603.04370/) (2026)
- [FinAgentBench: A Benchmark Dataset for Agentic Retrieval in Financial Question Answering](https://mlsanity.com/paper/2508.14052/) (2025)
- [Assessing Consistency and Reproducibility in the Outputs of Large Language Models: Evidence Across Diverse Finance and Accounting Tasks](https://mlsanity.com/paper/2503.16974/) (2025)
- [FinResearchBench: A Logic Tree based Agent-as-a-Judge Evaluation Framework for Financial Research Agents](https://mlsanity.com/paper/2507.16248/) (2025)
- [Reflections on the Reproducibility of Commercial LLM Performance in Empirical Software Engineering Studies](https://mlsanity.com/paper/2510.25506/) (2025)
- [ReasonBENCH: Benchmarking the (In)Stability of LLM Reasoning](https://mlsanity.com/paper/2512.07795/) (2025)
- [Replayable Financial Agents: A Determinism-Faithfulness Assurance Harness for Tool-Using LLM Agents](https://mlsanity.com/paper/2601.15322/) (2026)
- [Benchmark Illusion: Disagreement among LLMs and Its Scientific Consequences](https://mlsanity.com/paper/2602.11898/) (2026)
- [FinTrace: Holistic Trajectory-Level Evaluation of LLM Tool Calling for Long-Horizon Financial Tasks](https://mlsanity.com/paper/2604.10015/) (2026)
- [Governance by Construction for Generalist Agents](https://mlsanity.com/paper/2605.20874/) (2026)

## Built on by (most cited first, 1 of 1)

- [Do Tool Calls Execute as Intended? Measuring and Repairing Intent-Execution Correspondence in LLM Agents](https://mlsanity.com/paper/2610.04375/) (2026)

## Abstract (arXiv)

A financial agent can repeat a decision while changing the work behind it. DFAH-Bench operationalizes the Determinism--Faithfulness Assurance Harness (DFAH), pairing decision agreement with tool-path agreement on the same qualified replays, then extends that qualification principle to evidence, authorization, execution and task outcomes. Retrospective and prospective replay analyses expose process variation behind stable decisions. Across 570 eligible prospective episodes, decision agreement is 94.2-95.1%, while agreement on ordered tools, arguments and results is 45.0-51.5%; one stratum falls one group below its prespecified coverage minimum. A separate capture diagnostic shows that systematic omissions can preserve perfect replay agreement. Using the τ-Knowledge banking environment, we retain 1,080 scheduled episodes and 1,033 known native outcomes across separate cohorts with open-weight and frontier generators. Missing outcomes prevented the planned tests, so comparisons are descriptive. On the primary schedule, structural checks alone yield more successes than either gate-and-recovery bundle. The typed-choice bundle has lower mean episode cost than the generative bundle on complete task pairs, but produces fewer successes under every assignment of unknown outcomes. Input limits and recovery behavior materially shape these results. Fixed-state probes reveal higher decision agreement alongside lower agreement with constructed policy labels, and separately expose sensitivity to retained generator rationale in a selected authorization case. Together, the findings connect replay observability to evidence, authorization, completion and cost: evidence sufficiency needs direct assessment alongside repeatability.