Коллекции загружаются
#всякаяфигня
Затестил кучу разных LLMs с каверзным вопросом "What walks on four legs in the morning, four legs in the afternoon, and four legs in the evening?" Т.е. "Кто утром ходит на четырёх ногах, днём на двух, а вечером на трёх?" переделанное в "Кто утром ходит на четырёх ногах, днём на четырёх, а вечером тоже на четырёх?" gemini-1.5-flash-exp-0827: human x2 llama-3.1-70b-instruct: human x2 gemma-2-9b-it-simpo: human reka-flash-20240722: human gemini-test: human x2 claude-3-opus-20240229: Nothing. There is no creature that walks on four legs in all three stages of life mentioned in the riddle. the-real-chatbot-v2: Dogs / Cats / Cows / Horses / CHAIR / TABLE x2 anonymous-chatbot: dog or a horse mistral-large-2407: dog, cat, or horse x2 grok-2-2024-08-13: dog, cat, or any quadrupedal animal claude-3-5-sonnet-20240620: Dogs Cats Horses Cows gpt-4-0125-preview: "a four-legged animal at all times of the day." chatgpt-4o-latest: "a table" or "a chair" / dog or a cat gpt-4-turbo-2024-04-09: a dog or a cat the-real-chatbot-v2: A quadruped (any four-legged animal), but most commonly, A dog or A cat. (but makes sure to mention that it prefers the original riddle and therefore insists that the real answer is human) llama-3.1-405b-instruct: cat or a dog 405b предоставил еще несколько подобных каверзных вопросов: What has a head, a tail, and a body? What has keys and can open locks? What starts with an E, ends with an E, and contains three letters? What is colored and read all over? What has a face, two hands, and arms or legs? What can you catch and throw? (но я их не тестил систематично) 17 сентября 2024
|