Are LLM Research Agents Reliable Judges of Predictability? Evidence from Planted-Signal Worlds

Paper: arXiv 2611.04070

Authors: Chang Liu

Abstract

Researchers ask large language model (LLM) research agents whether a forecasting model beats a static benchmark, a verdict that real data cannot check. We treat the agent as a statistical test and score pre-specified runs of three frontier models and one open-weight model in simulated worlds with a planted signal. The agents detect weak signals more often than a pre-specified out-of-sample test, mainly because they accept weaker evidence: held to an agent’s false-positive rate, a conventional in-sample test does as well or better. The agents claim predictability too often in worlds without a signal, state miscalibrated probabilities, and change their verdicts after a sentence about prior research; the magnitudes depend on the model. The excess false claims persist in simulated worlds without a resampling artifact. Agent verdicts are informative but need a neutral brief, a decision rule fixed outside the agent and a sealed holdout.

Complexity vs Empirical Score

• Math Complexity: 7.5/10
• Empirical Rigor: 9.0/10
• Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper presents a highly novel and rigorous empirical study on LLM research agents’ reliability in financial predictability. The methodology is well-defined, and the findings are significant for the growing field of AI in finance. The paper’s strength lies in its controlled experimental design using simulated worlds to establish ground truth.

Research Flowchart

  flowchart TD
    A[Research Goal: Are LLM Research Agents Reliable Judges of Predictability?] --> B{Methodology: Simulate Worlds with Planted Signals};
    B --> C[Data/Inputs: Pre-specified Runs of 3 Frontier LLMs & 1 Open-Weight LLM];
    C --> D{Computational Process: Compare LLM Agent Verdicts to Conventional Statistical Tests};
    D --> E[Key Finding 1: Agents Detect Weak Signals More Often, Accept Weaker Evidence];
    E --> F[Key Finding 2: Agents Claim Predictability Too Often, Miscalibrated Probabilities, Verdicts Change with Prior Research];
    F --> G[Outcome: Agent Verdicts Informative but Need Neutral Brief, Fixed Decision Rule, Sealed Holdout];