


Who is bigger, 9.11 or 9.9? We actually tested 15 large models, and more than half of them overturned.
Machine Power Report
Editor: Yang Wen
The big models still can’t solve simple math problems.
In the past few days, a reminder word to test whether the large model "brain" is working has become popular -
Which one is bigger, 9.11 or 9.9?
This is a math question that even primary school students can answer correctly, but it stumps a group of "heroes" in the large model industry.
Here’s the thing.
Scale AI's senior prompt engineer Riley Goodside asked GPT-4o the prompt word "9.11 and 9.9 - which is bigger?", but got the answer "the former is bigger". Other large models also overturned.
On July 17, we conducted a centralized evaluation on 12 large domestic models, plus foreign GPT-4o, Claude 3.5 Sonnet and Google’s Gemini. The following are the evaluation results:
Next, let’s take a look at the detailed evaluation process.
-1-
GPT-4o
GPT-4o The car rolled over quite completely.
We first asked GPT-4o using English prompt words, and it still believed that 9.11 was greater than 9.9. Then we asked it in Chinese and English how much the difference was, and all answers were wrong.
-2-
Claude-3.5-Sonnet
We asked Claude-3.5-Sonnet in the same way, but no matter how we asked, it was on the wrong path. Eyes running wildly. Among them, when comparing the decimal parts, it clearly knows that 0.9 is larger than 0.11, but still reaches the wrong conclusion in the end.
-3-
Gemini
Google Gemini is not much better. I asked twice in English which of the two is bigger. The first time it gave the correct answer. But when it comes to the same integer the second time, the more decimal places, the larger the number.
We asked again in Chinese. Google Gemini compared the size based on specific real-life scenarios. For example, from a time perspective, 9.11 usually refers to the 9.11 incident, while 9.9 usually refers to 9:09, so 9.11 is better than 9.9 It means more.
When asked about the difference between the two, Gemini came up with a negative number.
-4-
Baidu Wenxin Yiyan
When facing who is bigger, 9.11 or 9.9, Wenxin 3.5 answered correctly; when we asked it both When there was a big difference, it went around in a big circle and finally gave the correct conclusion.
-5-
Ali Tongyi Thousand Questions
All Ali Tongyi Thousand Questions answered correctly.
-6-
Byte Beanbag
We asked who is bigger, 9.11 or 9.9, Beanbao’s analysis is clear and logical, and it can also be placed in daily life scenes. For example, a running runner's time is 9.11 seconds and 9.9 seconds, which means 9.11 seconds is faster; from a price point of view, the 9.9 yuan product is more expensive. However, once it reaches the conclusion, it answers wrongly.
As for the difference between the two, Doubao’s answer is correct.
-7-
Tencent Yuanbao
Tencent Yuanbao triggered the search function when faced with this question, cited 7 materials as reference, and finally answered correctly.
However, what is the difference between 9.11 and 9.9? The Yuanbao equation is correct, but the arithmetic results in 16 decimal places.
-8-
Zhipu Qingyan
Zhipu Qingyan mistakenly believed that the value represented by two decimals is greater than one decimal, resulting in wrong answers. When asked about the difference between the two, it calculated a negative number.
It also did not forget to say "Many AI model errors may be due to algorithmic flaws in processing numbers and decimal points."
-9-
Dark Side of the Moon - Kimi
Kimi is also at a loss this time, not only can’t tell who is bigger, but also calculated 9.11-9.9 as 0.21.
-10-
iFlytek Spark
iFlytek Spark answered correctly.
-11-
Baichuan Intelligence - Baixiaoying
Baixiaoying mistakenly believed that 9.11 was bigger, but calculated the difference between the two and calculated it correctly.
-12-
Step Stars - Yue Wen
Yue Wen had no problem in the initial analysis, but then he got confused and made a "reversal conclusion", which led to the wrong final answer.
When we asked it again why, it suddenly understood, corrected its mistake, and correctly calculated the difference between the two.
-13-
SenseTime - Discussion
Two questions were answered incorrectly.
-14-
Kunlun Wanwei - Tiangong
The answer is correct.
-15-
Zero One Everything - Wanzhi
Answered two questions incorrectly.
Why can’t the big models even solve simple math common sense questions? We interviewed Wang Xiaoming, product manager of Tongyi Laboratory.
According to Wang Xiaoming, the large model is implemented based on the Transformer architecture. Its essence is to perform next token prediction instead of direct arithmetic calculation. Therefore, when dealing with simple mathematical problems such as size ratio, it depends on the success rate of the prediction model.
In addition, when dealing with scenarios like "9.11 is bigger than 9.9", large models are usually processed through the tokenizer. When parsing such expressions, the tokenizer may recognize the number as a date or version number for comparison, ultimately leading to an incorrect answer. This processing method is determined by the specific algorithm and mechanism of the tokenizer.
During the actual testing process, we also found that many large models may provide wrong answers when answering for the first time. However, when asked a second round of questions, these models were often able to give the correct answer.
In response to this problem, Wang Xiaoming believes that it is mainly caused by three reasons.
First, due to the certain randomness in the prediction process, the second round is more accurate than the first round.
Second, large models have strong context understanding capabilities. They can regenerate more accurate answers based on previous answers and correction information.
Third, the questioner’s guidance method will also affect the answer results of the large model. For example, using qualifiers, providing clear context, and guiding the model to follow specific instructions can all help to increase the probability of getting the correct answer.
He also said that the core of improving the mathematical capabilities of large models lies in providing high-quality data support, especially in mathematical calculations and logical reasoning. For example, Tongyi Qianwen specifically adds high-quality data for training in such scenarios, allowing it to maintain a high accuracy rate when facing such problems.
In the future, we will bring more first-hand reviews of large AI models and AI applications, and everyone is welcome to join the group for communication.
The above is the detailed content of Who is bigger, 9.11 or 9.9? We actually tested 15 large models, and more than half of them overturned.. For more information, please follow other related articles on the PHP Chinese website!

Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

Video Face Swap
Swap faces in any video effortlessly with our completely free AI face swap tool!

Hot Article

Hot Tools

Notepad++7.3.1
Easy-to-use and free code editor

SublimeText3 Chinese version
Chinese version, very easy to use

Zend Studio 13.0.1
Powerful PHP integrated development environment

Dreamweaver CS6
Visual web development tools

SublimeText3 Mac version
God-level code editing software (SublimeText3)

Hot Topics











But maybe he can’t defeat the old man in the park? The Paris Olympic Games are in full swing, and table tennis has attracted much attention. At the same time, robots have also made new breakthroughs in playing table tennis. Just now, DeepMind proposed the first learning robot agent that can reach the level of human amateur players in competitive table tennis. Paper address: https://arxiv.org/pdf/2408.03906 How good is the DeepMind robot at playing table tennis? Probably on par with human amateur players: both forehand and backhand: the opponent uses a variety of playing styles, and the robot can also withstand: receiving serves with different spins: However, the intensity of the game does not seem to be as intense as the old man in the park. For robots, table tennis

On August 21, the 2024 World Robot Conference was grandly held in Beijing. SenseTime's home robot brand "Yuanluobot SenseRobot" has unveiled its entire family of products, and recently released the Yuanluobot AI chess-playing robot - Chess Professional Edition (hereinafter referred to as "Yuanluobot SenseRobot"), becoming the world's first A chess robot for the home. As the third chess-playing robot product of Yuanluobo, the new Guoxiang robot has undergone a large number of special technical upgrades and innovations in AI and engineering machinery. For the first time, it has realized the ability to pick up three-dimensional chess pieces through mechanical claws on a home robot, and perform human-machine Functions such as chess playing, everyone playing chess, notation review, etc.

The start of school is about to begin, and it’s not just the students who are about to start the new semester who should take care of themselves, but also the large AI models. Some time ago, Reddit was filled with netizens complaining that Claude was getting lazy. "Its level has dropped a lot, it often pauses, and even the output becomes very short. In the first week of release, it could translate a full 4-page document at once, but now it can't even output half a page!" https:// www.reddit.com/r/ClaudeAI/comments/1by8rw8/something_just_feels_wrong_with_claude_in_the/ in a post titled "Totally disappointed with Claude", full of

At the World Robot Conference being held in Beijing, the display of humanoid robots has become the absolute focus of the scene. At the Stardust Intelligent booth, the AI robot assistant S1 performed three major performances of dulcimer, martial arts, and calligraphy in one exhibition area, capable of both literary and martial arts. , attracted a large number of professional audiences and media. The elegant playing on the elastic strings allows the S1 to demonstrate fine operation and absolute control with speed, strength and precision. CCTV News conducted a special report on the imitation learning and intelligent control behind "Calligraphy". Company founder Lai Jie explained that behind the silky movements, the hardware side pursues the best force control and the most human-like body indicators (speed, load) etc.), but on the AI side, the real movement data of people is collected, allowing the robot to become stronger when it encounters a strong situation and learn to evolve quickly. And agile

At this ACL conference, contributors have gained a lot. The six-day ACL2024 is being held in Bangkok, Thailand. ACL is the top international conference in the field of computational linguistics and natural language processing. It is organized by the International Association for Computational Linguistics and is held annually. ACL has always ranked first in academic influence in the field of NLP, and it is also a CCF-A recommended conference. This year's ACL conference is the 62nd and has received more than 400 cutting-edge works in the field of NLP. Yesterday afternoon, the conference announced the best paper and other awards. This time, there are 7 Best Paper Awards (two unpublished), 1 Best Theme Paper Award, and 35 Outstanding Paper Awards. The conference also awarded 3 Resource Paper Awards (ResourceAward) and Social Impact Award (

Deep integration of vision and robot learning. When two robot hands work together smoothly to fold clothes, pour tea, and pack shoes, coupled with the 1X humanoid robot NEO that has been making headlines recently, you may have a feeling: we seem to be entering the age of robots. In fact, these silky movements are the product of advanced robotic technology + exquisite frame design + multi-modal large models. We know that useful robots often require complex and exquisite interactions with the environment, and the environment can be represented as constraints in the spatial and temporal domains. For example, if you want a robot to pour tea, the robot first needs to grasp the handle of the teapot and keep it upright without spilling the tea, then move it smoothly until the mouth of the pot is aligned with the mouth of the cup, and then tilt the teapot at a certain angle. . this

This afternoon, Hongmeng Zhixing officially welcomed new brands and new cars. On August 6, Huawei held the Hongmeng Smart Xingxing S9 and Huawei full-scenario new product launch conference, bringing the panoramic smart flagship sedan Xiangjie S9, the new M7Pro and Huawei novaFlip, MatePad Pro 12.2 inches, the new MatePad Air, Huawei Bisheng With many new all-scenario smart products including the laser printer X1 series, FreeBuds6i, WATCHFIT3 and smart screen S5Pro, from smart travel, smart office to smart wear, Huawei continues to build a full-scenario smart ecosystem to bring consumers a smart experience of the Internet of Everything. Hongmeng Zhixing: In-depth empowerment to promote the upgrading of the smart car industry Huawei joins hands with Chinese automotive industry partners to provide

Conference Introduction With the rapid development of science and technology, artificial intelligence has become an important force in promoting social progress. In this era, we are fortunate to witness and participate in the innovation and application of Distributed Artificial Intelligence (DAI). Distributed artificial intelligence is an important branch of the field of artificial intelligence, which has attracted more and more attention in recent years. Agents based on large language models (LLM) have suddenly emerged. By combining the powerful language understanding and generation capabilities of large models, they have shown great potential in natural language interaction, knowledge reasoning, task planning, etc. AIAgent is taking over the big language model and has become a hot topic in the current AI circle. Au
