Which table format do LLMs understand best?

https://www.improvingagents.com/blog/best-input-data-format-for-llms

by oidar • 10 months ago

230 110 10 months ago

ggm• 10 months ago

I find this extremely surprising. I would have expected dict structures to have higher semantic context associated with them.

svachalek2• 10 months ago

They did. The KV-Markdown is essentially a dict with ``` wrapper, and INI which is similar scored very high as well. The worst performers were index-based rows like CSV or Markdown tables. JSON is in the middle with high context and more syntactic noise and less clear record labels.

The odd ones to me are HTML which uses th and td to make indexed-based rows but did better than JSON somehow, and XML which is like JSON with even more syntactic noise placing better than INI. If I had to guess I'd say because vast amounts of the web were in the training set.

nightshift1• 10 months ago

I am not an expert on the subject but i suggest that you can also save context space by using shorter XML element names (like f instead of function, c instead of class, etc.). Just add a legend at the top or bottom to explain what each abbreviation means, LLMs can figure out the mapping without issues. I use this approach when generating project structure maps with Tree-sitter. I did a quick comparison and didn't notice much degradation with claude, so the context space you save may make it worthwhile. I would be interested to see a proper comparison.

1aurent29• 10 months ago

Common enough words like `function` and `class` are generally encoded as a single token by the tokenizer and may provide a slightly better context to the LLM. For openai you can test this stuff at https://platform.openai.com/tokenizer

Yiin• 10 months ago

if both f and function uses 1 token, are you really saving anything?

reidgreer• 10 months ago

interesting. I'm curious how this compares across different model families.

Sharlin• 10 months ago

where accuracy is paramount
accuracy: 60%

Not to mention that the least poorly performing format is probably the stupidest way to encode tabular data, beating even XML. But I guess that’s the new normal because we’re trying to shoehorn conversational AI models to every use case rather than, say, training finetunes that are better at particular tasks. (Yes, of course you can’t train finetunes when the model is a proprietary black box on someone else’s computer.) Something about hammers and nails…

mritchie712• 10 months ago

they used GPT-4.1 nano, results would be quite different with sonnet or gpt5.

lyu07282• 10 months ago

Or just regular gpt-4.1, it's a quite capable model.

fnordpiglet• 10 months ago

I was looking for the frontier curve where they tested their benchmark across different models since this sort of behavior is highly parameter, architecture, training, and fine tuning sensitive. It’s a practically useful question so I was really disappointed when a) they didn’t publish their code so you could test yourself, b) they didn’t do even a cursory examination of other models and sizes.

phyzome• 10 months ago

trust me bro, the next model bro, it's just way better bro

gpt5• 10 months ago

Isn't the best performing (markdown tables) and the worst (pipe delimited tables) basically the same format?

simonw3• 10 months ago

The best performing isn't markdown tables, it's markdown key/value pairs:

  ## Record 1
  
  ```
  id: 1
  name: Charlie A0
  age: 56
  city: New York
  department: Operations
  salary: 67896
  years_experience: 7
  project_count: 1
  ```

Which makes sense to me because the problem with formats like CSV and regular markdown tables is that it is too easy for the model to mistakenly associate a value in a row with the wrong header.

Explicit key/value formats like this or YAML or JSON objects make that a lot less likely.

mattcollins6• 10 months ago

I'm the person who ran the test.

To explain the 60% a bit more...

With small amounts of input data, the accuracy is near 100%. As you increase the size of the input data, the accuracy gradually decreases.

For this test, I intentionally chose an input data set large enough that the LLM would score in the region of 50% accuracy (with variation between formats) in order to maximise the discriminative power of the test.

ysleepy• 10 months ago

Wouldn't it be more useful to measure the number of rows the model can process while still hitting 100% accuracy?

rovr1382• 10 months ago

As you increase the size of the input data, the accuracy gradually decreases.

Interesting.

On your section "Limitations and Areas for Further Study",

What I'd be curious on future work would be,

    - changing the order of the data on each table type
    - changing the order of the questions

I'm curious to know if what it fails is the same, if it changes depending on the location, if it's a bias.

Is it always a specific question? Is it always a specific value? Is it always question #x (or around question #x?). Does it tend towards x or y on types of questions?

Good idea

padolsey2• 10 months ago

Thanks for your work on this! It's a very legit domain of problem for LLMs to optimize for. I've produced a comprehensive eval based on your post and run it against 30 models, each tasked with recalling specific data from 500 rows in different tabular formats. Have a look at the results here: https://weval.org/analysis/table-format-sensitivity__combine...

As you can see it's near 100% recall across all formats for a good chunk of frontier models, with a few (curiously, mostly Claude) failing a basic prompt adherance ("Return just the number") but still returning the right answers. The major failures are from Mistral Medium, Llama Maverick, Llama 3 70b Instruct, Mistral Nemo, Gemma 3 12b It, GPT 4o/4.1 Mini etc.

Based on these limited tests, here's the leaderboards on formats FWIW:

    CSV: 84.25%
    Markdown Table: 82.65%
    YAML: 81.85%
    JSON Lines (jsonl): 79.85%
    Markdown key-value: 79.83%
    Pipe-delimited: 79.45%
    Natural language summary: 78.65%
    JSON: 77.73%
    HTML table: 75.80%
    XML: 73.80%

So, the biggest takeaway really is: Use the best model you can reasonably afford, then format will matter less. The cheapest 100% coverage models are Gemini 2.5 Flash and Deepseek Chat V3.1

And if you have no control over model, then use CSV or Markdown Table.

Redster• 10 months ago

Thank you for including the tokens needed for each test.

It looks to me that the concisest way of representing each of these tables was a CSV and then a standard markdown table. The amount of tokens appears to be 1/2 or 1/3 of the other options. For experiments not in mice (GPT-4.1-nano), but in larger models or larger context aside from the data table itself, my guess is that preserving context is might be higher value than having the higher-LLM-legibility of the Markdown-KV.

rcarmo• 10 months ago

Hmmm. I’ve been using YAML data for tables for a while now, and had pretty good results.

cjonas• 10 months ago

The test really needed to be run on multiple data sizes (50, 100, 500, 1000, 5000). The more token efficient formats would probably eventually overtake the token heavy ones due to context pollution. All this test really says is what performs best for 1 particular model at one particular context length.

lmeyerov• 10 months ago

That's a cool concept - would be curious about a more common setup for agentic data analysis (ex: for using in Claude Code) like:

* Multiple tasks vs 1

* O3/o3-mini + 4o/4o-mini instead of nano

* Extra credit: Inside a fixed cost/length reasoning loop

Ex: does the md-kv benefit disappear with smarter models that you'r typically use, and thus just become a 2-3x cost?

brap• 10 months ago

I wonder how this compares to a more agentic approach where the LLM composes SQL queries to answer the questions, for example.

efitz2• 10 months ago

This was exactly my thought. Rather than feed the table directly to the LLM, build agents that extract the data and have the LLM act on the extracted data items. Then it’s a preference issue.

The author didn’t see much more than 60% accuracy which is not very useful for many (most?) real world tasks.

coeneedell2• 10 months ago

“Agents that extract the data” Are we really reinventing data frame readers to have an LLM in the critical path?

jitl• 10 months ago

Yeah I mean for many real world scale datasets you don’t want to blow the whole context window on a massive markdown file. Instead you can provide a tool that presents the data as a SQLite database. In my testing Claude code seems very capable of answering questions via SQLite queries or even `head` and `grep` on CSV files.

bwestergard• 10 months ago

But the result from the SQL query is going to be... a table. So at some point, tables need to go into context, and we need to know how well LLMs can incorporate those tables.

thom• 10 months ago

Well, ironically you then have the issue of how to present your database schema (including important things like the values in some categorical fields) to the LLM and in what format, so you never really escape this issue.

xnx5• 10 months ago

Title says "LLMs" (plural) but they only tested one

We only tested OpenAI’s GPT-4.1 nano.

picardo• 10 months ago

This should be higher. While the research question is interesting, the sample size makes the conclusion highly suspect. I'd like to see more research on this.

cwyers• 10 months ago

And not even a commonly used one. Gemini Flash or o4-mini would have been a much better choice if they wanted a cheap model

secwang• 10 months ago

maybe be org table

sega_sai• 10 months ago

Bizarre conclusions when on average all the formats perform poorly with average accuracy of 50%. Sure 60% is better than 40% but they are both unusable if you actually care about numbers...

zeitgeistcowboy• 10 months ago

My sentiments exactly. All the formats were so poorly read that they are all effectively useless.

zaidf• 10 months ago

I've been stunned by how many smart people talk so casually about LLMs becoming better at math. Do they just forget that a calculator that is wrong 1% of the time is a de facto calculator that doesn't work and should not be used?

xnx5• 10 months ago

I've been stunned by how many smart people talk so casually about LLMs becoming better at math

Could they be referring to this?

"Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad" https://deepmind.google/discover/blog/advanced-version-of-ge...

westoncb• 10 months ago

Doing math is not the same as calculating. LLMs can be very useful in doing math; for calculating they are the wrong tool (and even there they can be very useful, but you ask them to use calculating tools, not to do the calculations themselves—both Claude and ChatGPT are set up to do this).

If you're curious, check out how mathematicians like Robert Ghrist or Terence Tao are using LLMs for math research, both have written about it online repeatedly (along with an increasing number of other researchers).

Apart from assisting with research, their ability on e.g. math olympiad problems is periodically measured and objectively rapidly improving, so this isn't just a matter of opinion.

magicalhippo• 10 months ago

The best math lecturers I had at university sucked at mental calculations. Some almost screwed up 2+2 on the blackboard.

Yes LLMs suck at calculating stuff. However they can manipulate equations and such, and sometimes impressively so.

crazygringo2• 10 months ago

You realize that when typing into a calculator, you probably hit a wrong key more than 1% of the time? Which is why you always type important calculations twice?

I've been stunned by how many smart people talk so casually about how because LLMs aren't perfect, they therefore have no value. Do they just forget that nothing in the world is perfect, and the values of things are measured in degrees?

mattcollins6• 10 months ago

I'm the person who ran the test.

To hopefully clarify a bit...

I intentionally chose input data large enough that the LLM would be scoring in the region of 50% accuracy in order to maximise the discriminative power of the test.

__mharrison__• 10 months ago

Can you expand on how you did this?

fancyfredbot• 10 months ago

This is an interesting theoretical exercise but please for the love of god don't actually use an LLM to search tabular data. This is a solved problem. Free software does this with 100% accuracy and insane efficiency.

ModernMech• 10 months ago

This is a really eye-popping example. Because here we have input text that is fully structured perfectly unambiguous (it was carefully designed that way!) and yet the LLM can't get all the information out of it. Yet people are using these tools to summarize unstructured text, assuming the summary will capture the most salient points. Well how is the LLM supposed to be good for that task, if it can't even summarize the dang XML document? They keep telling me this thing is more expert than all the experts combined.

ComputerGuru• 10 months ago

Inputs were not long enough to properly see either of the true wins in terms of reduced token counts for terser formats or their benefits in terms of avoiding stuffing the context window thereby potentially reducing accuracy. The test really needs to be conducted across multiple dimensions!

dctoedt• 10 months ago

KSON? (I'm a complete ignoramus in this area but recently read about KSON in a piece posted here at HN.)

https://ochagavia.nl/blog/configuration-files-are-user-inter...

https://news.ycombinator.com/item?id=45291858 (135 comments)

freehorse• 10 months ago

Tbh I am more interested in processing data and formatting it to tabular forms than extracting data from tabular forms. One of the main uses I see in LLMs is structuring unstructured/semistructured data. I may occasionally feed a table to an LLM and ask such kinds of questions when I feel lazy, but I see no serious application of this as compared with using whatever language/library to process the data from the table (whether using an llm or not in the whole process). The point of having structured data is exactly this. But much more often I feed data to an llm and ask it to create a table.

veryrealsid• 10 months ago

I'm surprised by the accuracy, in practice, I feel like I generally have a lot better results

coeneedell2• 10 months ago

Do you measure your results in a repeatable way? In a way where your hypotheses about accuracy are falsifiable? Or do they just “feel” right?

mattcollins6• 10 months ago

I'm the person who ran the test.

The context I used in the test was pretty large. You'll see much better (near 100%) accuracy if you're using smaller amounts of context.

[I chose the context size so that the LLM would be scoring in the ballpark of 50% accuracy (with variation between formats) to maximise the discriminative power of the test.]

mingtianzhang• 10 months ago

The current OCR approach typically relies on a Vision-Language Model (VLM) to convert a table into a JSON structure. However, a table inherently has a 2D spatial structure, while Large Language Models (LLMs) are optimized for processing 1D sequential text. This creates a fundamental mismatch between the data representation and the model’s input format.

Most existing pipelines address this by preprocessing the table into a linearized 1D string before passing it to the LLM — a question-agnostic step that may lose structural information.

Instead, one could retain the original table form and, when a question is asked, feed both the question and the original table (as an image) directly into the VLM. This approach allows the model to reason over the data in its native 2D domain, providing a more natural and potentially more accurate solution.

fragmede2• 10 months ago

Yeah, I wonder how PNG would fare in this contest.

dcre2• 10 months ago

Only testing GPT-4.1-nano makes this basically useless. Most people are almost certainly using GPT-5 mini or better. This very poor analysis is like an LLM literacy test for readers.

grey-area3• 10 months ago

Please go away and do the work for us and let us know what anmazing accuracy you got with whatever version you think is better.

Anything below 100% is actually pretty useless when it comes to stats.

simonw3• 10 months ago

If you want 100% accuracy from these kinds of tasks with LLMs you can get it today, but you need to provide the LLM with the ability to run Python code and tell it to use something like Pandas.

You can confirm it's doing the right thing by reviewing the code it wrote.

dcre2• 10 months ago

Simon is right about using code execution, but many tables one might look at outside of formal data work are small enough for LLMs to be very reliable at, so this format question is practically relevant. I wish they had tested better models.

grey-area3• 10 months ago

They don’t understand any table formats; as shown by these results.

They can transform information in tables but information is lost due to that lack of understanding.

xnx5• 10 months ago

Great idea. Very limited execution. If they release the source data and question set, I'll repeat with more LLMs to flesh out the findings.

Ciantic• 10 months ago

This is a bit silly way to use LLMs to process tabular data. In reality, you'd ask it to write functions and execute them. First you'd ask it to create a type definition from the table, then ask it to create functions to process the data.

"Write a function to find years of experience by name? Return just the number, e.g. '12'."

It works much better, and it can single-shot many of the processing requirements just from type definitions it can infer from the data.

This way it's easier to stick to tabular formats that have easy reading libraries, like with TypeScript/JavaScript JSON, and with Python, maybe CSV...

sails• 10 months ago

I’d be interested in testing different data formats when using the structured outputs api

skyfantom• 10 months ago

Super surprised, I would expect CSV to beat all the others. And Markdown KV is something I hear first time about.

Bolwin• 10 months ago

It's made up, not a standard format

SweetSoftPillow• 10 months ago

Misleading title, just one LLM was tested.

jcheng5• 10 months ago

I 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 OPENAI_API_KEY, `uv sync`, and then run:

    uv run inspect eval evals/table_formats_eval.py@table_formats_markdown_kv --model openai/gpt-5 --limit 100

Update: Also, number of rows makes a massive difference, unsurprisingly; at 100 rows, gpt-4.1-nano scores 95%+ for both markdown-kv and csv. Both model and record count seem to matter a lot more than format.

jcheng5• 10 months ago

gpt-5 also got 100/100 for both CSV and JSON.

    uv run inspect eval evals/table_formats_eval.py@table_formats_csv --model openai/gpt-5 --limit 100
    uv run inspect eval evals/table_formats_eval.py@table_formats_json --model openai/gpt-5 --limit 100

xnx5• 10 months ago

Cool tool. I tried a few different things to get to work with google/gemini-2.5-pro, but couldn't figure it out.

jcheng5• 10 months ago

    uv add google-genai
    uv run scripts/run_benchmarks.py --models google/gemini-2.5-pro --formats markdown_kv --limit 100

And add GOOGLE_API_KEY=<your-key-here> to a file called .env in the repo root.

Unfortunately I started getting "quota exceeded" almost immediately, but it did give 6/6 correct answers before it crapped out.

fragmede2• 10 months ago

how about PNG?

catlifeonmars2• 10 months ago

Curious: how many iterations did you run of each benchmark and what was the variance?

solsane• 10 months ago

Did it make tool calls e.g. write code to read them?

jcheng5• 10 months ago

No, I didn’t provide any tools

lowbloodsugar• 10 months ago

In mice.

Or in this case gpt-4.1-nano

mattcollins6• 10 months ago

Author here.

This has made me chuckle several times - thanks!

olliem36• 10 months ago

We ended up making middleware for LLM 'tools/functions' that take common data/table formats like CSV, Excel and JSON.

The tool uses an LLM to write code to parse the data and conduct the analysis to return back to the LLM. Otherwise, we found pumping raw table data into a LLM is just not reliable, even if you go to the effort to conduct analysis on smaller chunks and merge the results.

jimjimjim• 10 months ago

accuracy: 60%

This should have been a python script.

How much of the current peak of the Gartner Hype Cycle should just be python scripts?

faxmeyourcode• 10 months ago

Curious how text-aligned tabular formats work for LLMs considering humans probably find them more readable than other formats

                                                                 System Sales(a)  
                                           Number of Units         (in Millions)  
         ──────────────────────────────────────────────────────────────────────── 
          KFC Division                              31,981    $           34,452  
          Taco Bell Division                         8,757                17,193  
          Pizza Hut Division                        20,225                13,108  
          Habit Burger & Grill Division                383                   713  
          YUM                                       61,346    $           65,466

I'm seeing pretty good success with extracting data out of 10-Qs which are formatted like this by default using the `edgartools` library's default `filing.text()` method.

johnfn• 10 months ago

The article has interesting data. But it’s frustrating to read AI generated text like this:

Performance Optimization: Reducing processing overhead while maintaining accuracy

What on earth does it mean that this “optimized performance”? This is nonsensical content. Performance wasn’t even measured, accuracy was. You can tell this was AI generated because “ Reducing processing overhead while maintaining accuracy” would likely be true for a perf optimization, but it has no meaning whatsoever in the context of the article.

This really throws into question whether I can take the rest of the article and data seriously.

wagslane• 10 months ago

I think they may be referring to token usage, which is mentioned in the article. fewer tokens = higher performance

naren87• 10 months ago

Great benchmark! It highlights an important but often downstream problem. In real-world pipelines, the bigger issue comes before this: extracting tables from PDFs or scans without breaking their layout. Once the structure is lost (merged headers, nested cells, footnotes, etc.), no data format can fully recover it.

Check out LLMWhisperer from Unstract —> it preserves table and layout fidelity when converting documents for LLM use. You can try it on complex PDFs or forms here: https://pg.llmwhisperer.unstract.com (no signup needed)

Layout preservation upstream often improves downstream accuracy more than choosing between CSV, JSON, or Markdown. Find more details here: https://unstract.com/llmwhisperer/

elliotto• 10 months ago

These sort of experiments and results are really important for language model implementation. This has a tangible implication for my AI startup and how we approach tool design.

Much more important than citation farming a paper on 1 % improved performance

noosphr• 10 months ago

I've found that xml is surprisingly good for llms when it comes to table extraction in production. I only found out when I send the raw xml storage format to benchmark again various flavours of everything else. XML turns out to the best format for tables that have more than three levels of nesting.

otterley• 10 months ago

Can someone explain why one would want to use an LLM to read tabular data? This is something even trivial code could do while using far fewer compute and energy resources.

101008• 10 months ago

Understanding the question is the hard part, that's where the LLM comes as an useful tool.

jcheng5• 10 months ago

I want this for after the code has run and returned results. Often when you use code to answer questions about a table, the result is in the form of a smaller table. I'd like to know how small that table needs to be before you can rely on the model being able to make reliable observations about it.

mitjam• 10 months ago

Was a bit surprised about the low csv performance, in my exp. it‘s very good (use it a lot with Excel and small tables, well below 100 rows).

As markdown kv performs so well, I am now curious about TOML.

hedgehog• 10 months ago

TOML works decently well both directions, useful if you need structured data out of models or APIs that don't support structured outputs.

beders• 10 months ago

60.7%

Why would anyone trust the output of an LLM, if it is barely better than guessing and much much worse than humans?

GPT-5 shows more impressive numbers, but for that particular task, the precision should be 100% - always. No matter how large the data set is or in which format. Why are we doing this?

padolsey2• 10 months ago

Interesting. Curious to reproduce across models, I made a comprehensive eval based on your post and ran it against 30 models, each tasked with recalling specific data from 500 rows in different tabular formats. Have a look at the results here: https://weval.org/analysis/table-format-sensitivity__combine...

As you can see it's near 100% recall across all formats for a good chunk of frontier models, with a few (curiously, mostly Claude) failing at basic prompt adherance ("Return just the number") but still returning the right answers. The major failures are from Mistral Medium, Llama Maverick, Llama 3 70b Instruct, Mistral Nemo, Gemma 3 12b It, GPT 4o/4.1 Mini etc.

Based on these limited tests, here's the leaderboards on formats FWIW:

    CSV: 84.25%
    Markdown Table: 82.65%
    YAML: 81.85%
    JSON Lines (jsonl): 79.85%
    Markdown key-value: 79.83%
    Pipe-delimited: 79.45%
    Natural language summary: 78.65%
    JSON: 77.73%
    HTML table: 75.80%
    XML: 73.80%

IMO the biggest takeaway really is: Use the best model you can reasonably afford, then the format chosen will matter less. The cheapest 100%-coverage models are Gemini 2.5 Flash and Deepseek Chat V3.1 FWIW. However, if you have no control over model, then use CSV or Markdown Table as these have highest chance of success.

The MAJOR issue that we might not want to admit is that there are a thousand confounders that prevent any meaningful canonical learning here. Crucially: The data within the tabular structure itself matters HUGELY. The scary probabilistic nature of LLMs mean the very subject of your queries can affect how the query is run, which is quite absurd from a IO/computing purity perspective. This is why tooling is so important. Enable the LLM to write and execute code safely, and you don't need to worry about such free-prose frailties.

hochstenbach• 10 months ago

There are other studies on this topic with similar results across LLM systems:

Y. Sui, M. Zhou, M. Zhou, S. Han, and D. Zhang, “Table Meets LLM: Can Large Language Models Understand Structured Table Data? A Benchmark and Empirical Study,” in Proceedings of the 17th ACM International Conference on Web Search and Data Mining, Merida Mexico: ACM, Mar. 2024, pp. 645–654. doi: 10.1145/3616855.3635752.

C. Pang, Y. Cao, C. Yang, and P. Luo, “Uncovering Limitations of Large Language Models in Information Seeking from Tables,” June 06, 2024, arXiv: arXiv:2406.04113. doi: 10.48550/arXiv.2406.04113.

TomBers• 10 months ago

Really interesting post. I ran into some of the limitations of working with tables and LLM's last year.

I experimented with an approach to use the llm to generate a bespoke transformation machine that uses an LLM to generate a series of transform steps to extracting key data from large data sets.

https://tombers.github.io/oblique-angles/ai/education/2025/0...

Rastonbury• 10 months ago

It appears that this is just testing data retrieval from somewhere in the table? Do the results translate to something where data analysis is performed? From something as simple as summing across rows or averages to generating graphs.

I once tried to get Claude and ChatGPT to build me a excel financial model, failed pretty hard. The models seem to lose track where they are in a table

pietz• 10 months ago

This article screams for a accuracy vs. tokens plot. Thanks though, interesting results.

mattcollins6• 9 months ago

Here you go: https://www.improvingagents.com/blog/best-input-data-format-...

BrokenLButton• 10 months ago

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.

selinkocalar• 10 months ago

We'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.

Which table format do LLMs understand best?

Related Stories