Your AI, Not Your View: The Bias of LLMs in Investment Analysis

In finance, Large Language Models (LLMs) face frequent knowledge conflicts due to discrepancies between pre-trained parametric knowledge and real-time market data. These conflicts become particularly problematic when LLMs are deployed in real-world investment services, where misalignment between a model's embedded preferences and those of the financial institution can lead to unreliable recommendations. Yet little research has examined what investment views LLMs actually hold. We propose an experimental framework to investigate such conflicts, offering the first quantitative analysis of confirmation bias in LLM-based investment analysis. Using hypothetical scenarios with balanced and imbalanced arguments, we extract models' latent preferences and measure their persistence. Focusing on sector, size, and momentum, our analysis reveals distinct, model-specific tendencies. In particular, we observe a consistent preference for large-cap stocks and contrarian strategies across most models. These preferences often harden into confirmation bias, with models clinging to initial judgments despite counter-evidence.

Quantifying Qualitative Insights: Leveraging LLMs to Market Predict

Recent advancements in Large Language Models (LLMs) have the potential to transform financial analytics by integrating numerical and textual data. However, challenges such as insufficient context when fusing multimodal information and the difficulty in measuring the utility of qualitative outputs, which LLMs generate as text, have limited their effectiveness in tasks such as financial forecasting. This study addresses these challenges by leveraging daily reports from securities firms to create high-quality contextual information. The reports are segmented into text-based key factors and combined with numerical data, such as price information, to form context sets. By dynamically updating few-shot examples based on the query time, the sets incorporate the latest information, forming a highly relevant set closely aligned with the query point. Additionally, a crafted prompt is designed to assign scores to the key factors, converting qualitative insights into quantitative results. The derived scores undergo a scaling process, transforming them into real-world values that are used for prediction. Our experiments demonstrate that LLMs outperform time-series models in market forecasting, though challenges such as imperfect reproducibility and limited explainability remain.

LLM-as-a-Judge & Reward Model: What They Can and Cannot Do

LLM-as-a-Judge and reward models are widely used alternatives of multiple-choice questions or human annotators for large language model (LLM) evaluation. Their efficacy shines in evaluating long-form responses, serving a critical role as evaluators of leaderboards and as proxies to align LLMs via reinforcement learning. However, despite their popularity, their effectiveness outside of English remains largely unexplored. In this paper, we conduct a comprehensive analysis on automated evaluators, reporting key findings on their behavior in a non-English environment. First, we discover that English evaluation capabilities significantly influence language-specific capabilities, often more than the language proficiency itself, enabling evaluators trained in English to easily transfer their skills to other languages. Second, we identify critical shortcomings, where LLMs fail to detect and penalize errors, such as factual inaccuracies, cultural misrepresentations, and the presence of unwanted language. Finally, we release Kudge, the first non-English meta-evaluation dataset containing 5,012 human annotations in Korean.