Open experiments and findings from Bean Labs — the Finance AI Agent research initiative by Beancount.io.
PHANTOM (NeurIPS 2025) is the first benchmark to measure LLM hallucination detection on real SEC filings across context lengths up to 30,000 tokens. Qwen3-30B-A3B-Thinking leads with F1=0.882; 7B models score near random guessing — with direct implications for autonomous accounting agents.
FinMaster (arXiv:2505.13533) benchmarks o3-mini, Claude 3.7 Sonnet, and DeepSeek-V3 across 183 financial tasks—revealing that models score 96% on financial literacy but collapse to 3% on statement generation, with multi-step consulting tasks losing 21 accuracy points from error propagation.
ReAct (Yao et al., ICLR 2023) interleaves chain-of-thought reasoning with tool actions in a single trajectory, outperforming pure CoT on fact verification and imitation learning on embodied tasks by 34 percentage points. This analysis covers the paper's failure modes — search-induced distraction and compounding errors — and what they mean for autonomous agents writing back to Beancount ledgers.
A close reading of Toolformer (Meta AI, NeurIPS 2023): how perplexity-filtered self-supervised training teaches a 6.7B-parameter model to call external APIs, where it outperforms GPT-3 175B on arithmetic benchmarks, and why its single-step architecture cannot support the chained tool calls required for structured ledger operations.
FinBen evaluates 15 LLMs across 36 financial datasets at NeurIPS 2024, finding GPT-4 reaches 0.63 Exact Match on numerical QA and 0.54 on stock movement forecasting — near chance. Here is what those numbers mean for building a reliable accounting agent on a Beancount ledger.
