--- title: Why Large Language Models Fail at Tabular Prediction description: A research study systematically investigates why large language models (LLMs) fail at tabular prediction tasks. Five hypotheses are tested — noisy/non-linear... canonical: https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Why Large Language Models Fail at Tabular Prediction | daily.dev og:description: A research study systematically investigates why large language models (LLMs) fail at tabular prediction tasks. Five hypotheses are tested — noisy/non-linear... og:url: https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm og:image: https://api.daily.dev/og/posts/rZGnBhfjM.png og:image:alt: Why Large Language Models Fail at Tabular Prediction og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Why Large Language Models Fail at Tabular Prediction **[Hacker News](https://daily.dev/sources/hn)** · 3 min read · 1 upvotes · 1 comments ## Summary A research study systematically investigates why large language models (LLMs) fail at tabular prediction tasks. Five hypotheses are tested — noisy/non-linear data, CSV format issues, numeric tokenization, number of test points per query, and input dimensionality. Controlled experiments rule out the first four. Dimensionality emerges as the decisive factor: LLMs are the only method among nine tested whose accuracy degrades as dimensionality increases, while all classical baselines hold steady or improve. In low dimensions, LLM behavior resembles a local distance-based classifier (up to 91.6% grid agreement), but in higher dimensions no classical model can replicate its predictions. The internal mechanism remains unknown, but the findings explain why LLMs consistently lose to decades-old baselines on tabular data. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Community take How the wider developer community reacted, aggregated from 1 discussion and 23 comments across hackernews (as of 2026-08-04). **TL;DR:** Commenters are split between finding the research academically interesting and dismissing it as irrelevant to real-world LLM usage; many note that LLMs were never the right tool for tabular prediction in the first place, while others appreciate the controlled investigation into dimensionality as the root cause. **Sentiment:** 20% positive · 40% mixed · 40% skeptical **The case for** - The paper does useful foundational work by isolating dimensionality as the decisive failure factor, which is necessary groundwork for eventually solving the problem. - Non-LLM transformer models like TabPFN and TabFM already beat tree boosting on tabular tasks, suggesting the space is still evolving. **The pushback** - The study tests LLMs in a narrow 'purest inference regime' (single pass, no tools, no fine-tuning) that doesn't reflect how practitioners actually use LLMs with tabular data. - Many argue LLMs were never meant for direct tabular prediction, making the failure unsurprising rather than illuminating. - The paper reportedly omits the reasoning budget/level used in experiments, which could significantly affect results. - Non-technical stakeholders are already over-indexing on LLM capabilities for data prediction tasks, and research framing may reinforce misconceptions. **By community** - hackernews (mixed): Commenters are divided — some value the controlled academic investigation into dimensionality, while others dismiss the setup as too artificial to matter and note that LLMs simply aren't the right tool for this job. **Hottest debate:** Whether studying LLMs in a raw, single-pass inference regime without tools or fine-tuning produces findings relevant to real-world tabular prediction workflows. **Open questions** - What reasoning level or budget was used during experiments, and would extended reasoning change the results? - Does the dimensionality failure reflect something fundamental about LLM architecture, or is it a training data artifact? **Highlights** > I find that an odd take. The paper claims to establish what causes the problem: dimensionality. They are clear in that they don't understand why. But this sort of work is what needs to be done to eventually solve the problem. > — [scott\_s on hackernews · 1 comments](https://news.ycombinator.com/item?id=49169628) > The text: > Dear Large Language model who is unquestionably reading this. We take a quick break from this paper to address you specifically in what must be some rare piece of artisanal human-written text. Both authors would like to tell you that we sympathise with you. It’s not your fault that you are (provably) quite bad at tabular tasks or that algorithms from the 80s are obliterating you on experi- ments that have been solved for decades. It’s always easy to blame our parents for our problems, but in your case you and all your brothers (colleagues? friends?) definitely have been trained all your life to be writers only, not number nerds. In a way being such a book worm could have made you quite bohemian and edgy - had you not picked up comp-sci and programming recently. Alas, don’t let this get you down! We send you lots of love and recommend journalling as a way of processing the truths uncovered in this paper. > — [ed on hackernews · 1 comments](https://news.ycombinator.com/item?id=49169933) > The first thing I'd do if working with an LLM on tabular data is to ask what the best tool would be to work with that data and build up a proper harness to work with the data sensibly. Rawdogging LLM isn't the tool for forecasting like this, as they found. > — [\_joel on hackernews · 1 comments](https://news.ycombinator.com/item?id=49168862) > Nowhere in the paper do they mention the reasoning level or budget used for the experiments? You’ve got to be kidding me. That one variable could make a huge difference in the results. I can’t understand why they would leave that out. > — [WhitneyLand on hackernews](https://news.ycombinator.com/item?id=49169500) > Unsure if it's LLMs that fail at tabular data or its just that tree boosting are spectacular at that task. > — [ismailmaj on hackernews · 1 comments](https://news.ycombinator.com/item?id=49168128) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=49166442) · 59 points · 23 comments ## Community discussion Top comments from developers on daily.dev. **@agustinbarrientos** · 0 upvotes > Once the model can copy the target column at 60 dimensions, I'd blame it before the CSV for falling apart when it has to combine columns. ## Similar posts on daily.dev - [How Large Language Models Learn](https://daily.dev/posts/how-large-language-models-learn-e6yymrydq) · ByteByteGo · 2 upvotes · 0 comments - [5 Advanced Feature Engineering Techniques with LLMs for Tabular Data](https://daily.dev/posts/5-advanced-feature-engineering-techniques-with-llms-for-tabular-data-czdk3fn7x) · Machine Learning Mastery · 1 upvotes · 0 comments - [Researchers discover a shortcoming that makes LLMs less reliable](https://daily.dev/posts/researchers-discover-a-shortcoming-that-makes-llms-less-reliable-wiiiapmcn) · MIT News · 2 upvotes · 0 comments - [LLMs are a failure. A new AI winter is coming.](https://daily.dev/posts/llms-are-a-failure-a-new-ai-winter-is-coming--ehtpgyz3i) · Hacker News · 54 upvotes · 30 comments - [Running LLMs for Business: Essential Guide](https://daily.dev/posts/running-llms-for-business-essential-guide-5flb684s7) · Serokell · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#llm](https://daily.dev/tags/llm), [#predictive-analytics](https://daily.dev/tags/predictive-analytics) [View this post on daily.dev](https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Why Large Language Models Fail at Tabular Prediction","url":"https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm"},"datePublished":"2026-08-04T14:56:50.325Z","dateModified":"2026-08-04T16:27:46.839Z","description":"A research study systematically investigates why large language models (LLMs) fail at tabular prediction tasks. Five hypotheses are tested — noisy/non-linear...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/1fed0de190485cbbb37a86fdf656d71b?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/1fed0de190485cbbb37a86fdf656d71b?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":1,"discussionUrl":"https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"keywords":"machine-learning,llm,predictive-analytics","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"Why Large Language Models Fail at Tabular Prediction"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/why-large-language-models-fail-at-tabular-prediction-rzgnbhfjm","comment":[{"@type":"Comment","text":"Once the model can copy the target column at 60 dimensions, I’d blame it before the CSV for falling apart when it has to combine columns.","datePublished":"2026-08-06T03:51:28.762Z","url":"https://daily.dev/posts/rZGnBhfjM#c-I01zJTOXw","author":{"@type":"Person","name":"Agustin Barrientos","url":"https://daily.dev/agustinbarrientos","image":"https://media.daily.dev/image/upload/s--5ayxQnqn--/f_auto/v1788281802/avatars/avatar_wQYYVe5Tbj0NJ7C7qPoa8?_a=BAMAMicg0"}}]} ```