LLMs have achieved performance that surpasses humans in many NLP tasks. However, it remains unclear whether problems that are difficult for humans are also difficult for LLMs. This study investigates how the difficulty of quizzes in a buzzer setting differs between LLMs and humans. Specifically, we first collect Japanese quiz data including questions, answers, and correct response rate of humans, then prompted LLMs to answer the quizzes under several settings, and compare their correct answer rate to that of humans from two analytical perspectives. The experimental results showed that, compared to humans, LLMs struggle more with quizzes whose correct answers are not covered by Wikipedia entries, and also have difficulty with questions that require numerical answers.

"I don't feel particularly unhappy about my work being used to train AI," says Japanese novelist Rie Qudan. "Even if it is copied, I feel confident there's a part of me that will remain, which nobody can copy." The 34-year old author is talking to me via Zoom from her home near Tokyo, ahead of the publication of the English-language translation of her fourth novel, Sympathy Tower Tokyo. The book attracted controversy in Japan when it won a prestigious prize, despite being partly written by ChatGPT. At the heart of Sympathy Tower Tokyo is a Japanese architect, Sara Machina, who has been commissioned to build a new tower to house convicted criminals. It will be a representation of what one character – not without irony – calls "the extraordinary broadmindedness of the Japanese people", in that the tower will house offenders in compassionate comfort.

The advancement of large language models has significantly improved natural language processing. However, challenges such as jailbreaks (prompt injections that cause an LLM to follow instructions contrary to its intended use), hallucinations (generating incorrect or misleading information), and comprehension errors remain prevalent. In this report, we present a comparative analysis of the performance of fifteen distinct models, with each model undergoing a standardized test comprising 38 queries across three key metrics: jailbreaks, hallucinations, and comprehension errors. The models are assessed based on the total occurrences of jailbreaks, hallucinations, and comprehension errors. Our work exposes these models' inherent vulnerabilities and challenges the notion of human-level language comprehension of these models. We have empirically analysed the impact of non-standard Unicode characters on LLMs and their safeguarding mechanisms on the best-performing LLMs, including GPT-4, Gemini 1.5 Pro, LlaMA-3-70B, and Claude 3 Opus. By incorporating alphanumeric symbols from Unicode outside the standard Latin block and variants of characters in other languages, we observed a reduction in the efficacy of guardrails implemented through Reinforcement Learning Human Feedback (RLHF). Consequently, these models exhibit heightened vulnerability to content policy breaches and prompt leakage. Our study also suggests a need to incorporate non-standard Unicode text in LLM training data to enhance the capabilities of these models.