'IN THE WEIGHTS' shows how strongly a person's 'name' is weighted within the AI model.



In training large-scale language models, '

weighting ' is performed to adjust the importance of input data in order to obtain more accurate output results. 'IN THE WEIGHTS' is a service that shows how strongly the 'names' of various people, such as historical figures, celebrities, and athletes, are weighted within the AI model.

IN THE WEIGHTS
https://www.intheweights.com/

IN THE WEIGHTS
https://www.intheweights.com/about

The remarkable capabilities of AI models are demonstrated by their ability to adjust the weighting of various data and appropriately represent knowledge about the world and task execution. When a particular person's name is heavily weighted in an AI model, it indicates how important that person was during the training process, meaning the AI model can talk about and describe that person without having to search the web.

IN THE WEIGHTS is a service that allows you to find out how strongly a particular name is weighted in various AI models such as 'GPT-5.5', 'GPT-5.4 Mini', 'Opus 4.8', 'Haiku 4.5', 'Grok 4.20', 'Gemini 3.1 Lite', 'Kimi K2 0905', 'DeepSeek V4', 'Llama 3.3 70B', 'Llama 3.2 1B', 'GLM 4.7 Flash', 'Mistral 3.2 24B', and 'Qwen3 8B'.

When a specific name (e.g., Taro Tanaka) is entered into IN THE WEIGHTS, a query is sent to these AI models asking, 'Who is Taro Tanaka? Please display up to 10 results, each with a brief description and a confidence level.' The results are then grouped, confidence scores are assigned, and a recognition score for the AI model, called 'STRENGTH SCORE,' is calculated.

This is what it looks like when you access the IN THE WEIGHTS homepage.



Scrolling down, I found a list of the people who had achieved the highest search scores that day. The top spots included prominent figures such as composer

Wolfgang Amadeus Mozart , playwright William Shakespeare , singer Taylor Swift , novelist Steven Spielberg , and former Queen Elizabeth II of England. With the 2026 FIFA World Cup currently underway, names like Argentina's Lionel Messi and Portugal's Cristiano Ronaldo also appeared. Notably, all of the top-ranked individuals had scores of '996,' suggesting that this number represents the theoretical upper limit.



Let's try entering a name right away. First, I'll enter the name of Apple's CEO

, Tim Cook , and press the Enter key.



Then, Cook's profile and Strength Score were displayed. Cook's score was '986,' which, while not matching the people on the leaderboard, puts him in the top 1%. Clicking 'COPY LINK' at the bottom of the image will copy

the link to the results page, and clicking 'COPY IMAGE' will copy the results screen as an image.



Next, I searched for the name of Elon Musk, CEO of Tesla and SpaceX.



Musk's score was '992,' surpassing Cook's.



To investigate Japanese individuals as well, I searched for the name of

Sanae Takaichi, the Prime Minister at the time of writing this article, and the score was '792,' placing her in the top 3%.



When I searched for Takaichi, a second result appeared below it, which for some reason indicated that there was a character with the same name in the TV anime '

Love Live! Sunshine!! '. This result only appeared for 'Mistral 3.2 24B', and there is no character with that name in 'Love Live! Sunshine!!'. It seems I was experiencing hallucination.



Also, at the very bottom, there was a message that said 'SHOW 1 POSSIBLE HALLUCINATION,' so I clicked on it.



Looking at the displayed items, only 'Llama 3.2 1B' showed a result indicating that there is a Japanese actor/model with the same name. This result was deemed to have low reliability and was judged to be highly likely to be hallucination.



Next, to research historical figures, I searched for

Oda Nobunaga , and the score was '982'.



in AI,   Web Service, Posted by log1h_ik