Markdown being the most understandable format lines up with my anecdotal experience. But I also was only using OpenAI models at the time. I do think there are a lot of unexplored methods/table transformation tools that we haven't even thought of yet.
Which table format do LLMs understand best?
111–115 of 115 posts
Re: Which table format do LLMs understand best?
#112I was curious enough to have Codex create a similar benchmark: https://github.com/jcheng5/table-formats With 1000 rows and 100 samples and markdown-kv, I got these scores: - gpt-4.1-nano: 52% - gpt-4.1-mini: 72% - gpt-4.1: 93% - gpt-5: 100% I was so surprised by gpt-5 getting 100% that I ran it again with 1000 samples. It got 999 correct, and one wrong. To reproduce it yourself, clone the repo, add a .env file with O…
Did it make tool calls e.g. write code to read them?
Re: Which table format do LLMs understand best?
#113We've experimented with different formats for feeding data to LLMs and markdown tables usually work pretty well. JSON is more structured but harder for the model to parse visually.
CSV works okay but you lose a lot of context about what the columns actually represent. The model performs better when it can 'see' the structure clearly.
Re: Which table format do LLMs understand best?
#114I was curious enough to have Codex create a similar benchmark: https://github.com/jcheng5/table-formats With 1000 rows and 100 samples and markdown-kv, I got these scores: - gpt-4.1-nano: 52% - gpt-4.1-mini: 72% - gpt-4.1: 93% - gpt-5: 100% I was so surprised by gpt-5 getting 100% that I ran it again with 1000 samples. It got 999 correct, and one wrong. To reproduce it yourself, clone the repo, add a .env file with O…
Did it make tool calls e.g. write code to read them?
No, I didn’t provide any tools
Re: Which table format do LLMs understand best?
#115This article screams for a accuracy vs. tokens plot. Thanks though, interesting results.