% ============================================================================
\section{Related Work}
\label{sec:literature}
% ============================================================================

VQuant sits at the intersection of four research strands: (i)~large language models and
reasoning, (ii)~tool-augmented and agentic systems, (iii)~retrieval augmentation and grounding,
and (iv)~the application of LLMs to finance. We review each in turn and position our
contribution accordingly.

\subsection{Language models and reasoning}
\label{subsec:lit_lms}

The transformer architecture of \citet{vaswani2017attention} and the demonstration of
in-context, few-shot learning at scale by \citet{brown2020language} established the modern
paradigm of general-purpose language models. A complementary line of work showed that
intermediate ``chain-of-thought'' reasoning improves performance on multi-step problems
\citep{wei2022chain}, motivating a generation of models that allocate explicit computation to
deliberation before answering. VQuant builds directly on this development: it targets
reasoning-augmented models that can ``think'' under an adaptive token budget prior to
responding, and it surfaces this intermediate reasoning to the user. We treat the foundation
model as a fixed, externally provided capability \citep{bommasani2021opportunities} and study
the \emph{system} that must be engineered around it.

\subsection{Tool use and agentic systems}
\label{subsec:lit_agents}

A model confined to its parametric memory cannot retrieve live prices or perform exact
arithmetic. The tool-use literature addresses this by teaching models to call external
functions. \citet{schick2023toolformer} showed that models can learn to invoke APIs, while the
ReAct framework of \citet{yao2023react} interleaved reasoning traces with actions in an
observe--think--act loop that has become the template for modern agents. Program-aided
approaches \citep{gao2023pal} delegate computation to a code interpreter, offloading exact
arithmetic from the language model to a deterministic runtime---a principle that VQuant adopts
as a first-class design constraint. Industrial function-calling protocols
\citep{anthropic2024tooluse} standardized the schema-driven invocation of tools, and the Model
Context Protocol generalized tool exposure across providers. VQuant operationalizes these ideas
at scale: rather than a handful of tools, it exposes a curated catalogue of more than two
hundred, and it contributes practical mechanisms---prompt-level routing, execution batching,
and a single-call computation paradigm---for managing the resulting orchestration complexity.

\subsection{Retrieval augmentation and grounding}
\label{subsec:lit_rag}

Retrieval-augmented generation \citep{lewis2020retrieval} grounds model outputs in external
documents, mitigating hallucination and staleness. VQuant generalizes grounding beyond text
retrieval: its ``retrieval'' surface includes structured market-data endpoints, semantic and
keyword web search, and on-demand numerical computation. Every quantitative claim in a VQuant
report is traceable to a deterministic source, which we view as the financial-domain analogue
of citation-grounded generation. This emphasis on auditability connects to broader concerns
about reliability and provenance in machine learning for finance \citep{lopezdeprado2018advances}.

\subsection{Large language models in finance}
\label{subsec:lit_finance}

Domain-adapted models such as BloombergGPT \citep{wu2023bloomberggpt} and open frameworks such
as FinGPT \citep{yang2023fingpt} demonstrated the value of finance-specific training data.
On the application side, \citet{lopezlira2023chatgpt} found that LLM-derived sentiment from
news headlines predicts subsequent returns, and \citet{kim2024financial} showed that LLMs can
analyze financial statements at a level competitive with human analysts. These studies largely
treat the LLM as a \emph{forecasting or classification} engine operating on text. VQuant takes
a different stance: it uses the LLM as an \emph{orchestrator} that assembles live data,
delegates numerical work to deterministic code, and composes the results into a report. The
classical quantitative methods it automates---option pricing \citep{black1973pricing,
merton1973theory}, conditional-volatility modeling \citep{engle1982autoregressive,
bollerslev1986generalized}, Monte Carlo valuation \citep{glasserman2004monte}, and
Value-at-Risk \citep{jorion2007value}---are well established; the contribution is the
\emph{architecture} that makes them accessible through natural language while preserving
reproducibility.

\subsection{Positioning}
\label{subsec:lit_positioning}

Relative to prior work, VQuant is distinctive in three respects. First, it integrates an
unusually broad and heterogeneous tool ecosystem under one protocol, and it contributes
concrete methodology for routing and rate-managing that ecosystem. Second, it enforces a strict
separation between reasoning (performed by the LLM) and computation (performed by deterministic
code), which we argue is the appropriate design for numerically sensitive domains. Third, it
treats the \emph{deliverable}---a citable, exportable, reproducible document---as a primary
artifact rather than an afterthought. The remainder of the paper develops each of these points.