Idea Ripper // ArchiveCards: 1555 [Live] | Updated: 2026-10-14
Card #c22IR // #c22Synthesis

a financial fine-tuning experiment is not defined by a model name and a training set. It is defined by a decision date, a permissible information set, a task loss, a calibration protocol, and an approval boundary.

— In our words · Source: Large Language Models in Finance (Miquel Noguer i Alonso)

[Use when]

Before greenlighting any fine-tuning run, demand the contract: what data as of when, what loss, what calibration check, who approved it.

[Why it matters]

Model name plus training set is how demos are described; decision date plus information set plus loss plus calibration plus approval boundary is how production systems are defended. Without the contract, a fine-tune can't be reproduced, audited, or blamed — it can only be admired.

Related principles

#c19A valid evaluation must freeze not only the price history but also the prompt, retrieval store, tools, memory, model version, and…Use when reviewing or designing a backtest of an LLM-based forecasting or trading system. #c23Two things distinguish it from a generic one, and both are controls rather than models. The availability filter D ≤ t admits only documents…When evaluating a RAG build, check the two controls first: does retrieval enforce the decision-time cutoff,…

More from Large Language Models in Finance

#c13Prediction and agency should not be conflated. A credit model can estimate the probability that a borrower breaches a covenant. An agentic…Use when a team proposes a stronger model as the fix for an agent that lacks memory, permissioning, or an… #c14The technically and institutionally mature objective is bounded autonomy : the agent can plan, retrieve, calculate, and propose actions…Use when deciding what an AI agent is permitted to do in production: which actions stay read-only, which need… #c15This principle prevents a common category error. A model that says it will respect a mandate has not yet respected the mandate. A model…Use when a vendor demo, benchmark score, or red-team chat log is being offered as proof that an agent is safe…