Evaluating AI for Finance: Is AI Credible at Assessing Investment Risk?

We evaluate the credibility of leading AI models in assessing investment risk appetite. Our analysis spans proprietary (GPT-4, Claude 3.7, Gemini 1.5) and open-weight models (LLaMA 3.1/3.3, DeepSeek-V3, Mistral-small), using 1,720 user profiles constructed with 16 risk-relevant features across 10 countries and both genders. We observe significant variance across models in score distributions and demographic sensitivity. For example, GPT-4o assigns higher risk scores to Nigerian and Indonesian profiles, while LLaMA and DeepSeek show opposite gender tendencies in risk classification. While some models (e.g., GPT-4o, LLaMA 3.1) align closely with expected scores in low- and mid-risk ranges, none maintain consistent performance across regions and demographics. Our findings highlight the need for rigorous, standardized evaluations of AI systems in regulated financial contexts to prevent bias, opacity, and inconsistency in real-world deployment.

Adapting Automatic Speech Recognition for Accented Air Traffic Control Communications

Effective communication in Air Traffic Control (ATC) is critical to maintaining aviation safety, yet the challenges posed by accented English remain largely unaddressed in Automatic Speech Recognition (ASR) systems. Existing models struggle with transcription accuracy for Southeast Asian-accented (SEA-accented) speech, particularly in noisy ATC environments. This study presents the development of ASR models fine-tuned specifically for Southeast Asian accents using a newly created dataset. Our research achieves significant improvements, achieving a Word Error Rate (WER) of 0.0982 or 9.82% on SEA-accented ATC speech. Additionally, the paper highlights the importance of region-specific datasets and accent-focused training, offering a pathway for deploying ASR systems in resource-constrained military operations. The findings emphasize the need for noise-robust training techniques and region-specific datasets to improve transcription accuracy for non-Western accents in ATC communications.

WalledEval: A Comprehensive Safety Evaluation Toolkit for Large Language Models

WalledEval is a comprehensive AI safety testing toolkit designed to evaluate large language models (LLMs). It accommodates a diverse range of models, including both open-weight and API-based ones, and features over 35 safety benchmarks covering areas such as multilingual safety, exaggerated safety, and prompt injections. The framework supports both LLM and judge benchmarking, and incorporates custom mutators to test safety against various text-style mutations such as future tense and paraphrasing. Additionally, WalledEval introduces WalledGuard, a new, small and performant content moderation tool, and SGXSTest, a benchmark for assessing exaggerated safety in cultural contexts. We make WalledEval publicly available at https://github.com/walledai/walledevalA.