---
format: "aidr-story-markdown/v1"
id: "1db9646a1ed86378655058e7036f18092316497fd5293f54eb40669206b550af"
canonical_url: "https://aidr.today/1db9646a?lang=en"
title: "TabPFN and TabICL vs. tuned XGBoost: the model that doesn't train won 14/14"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-28T02:27:40.000Z"
category: "Research"
topics: ["inference"]
source_urls: ["https://efraingaray.com/en/blog/tabpfn-vs-xgboost/","https://news.ycombinator.com/item?id=49872864"]
summary: "The claim behind TabPFN and TabICL is that they predict on a table without ever training on it and still beat tuned boosting."
---

# TabPFN and TabICL vs\. tuned XGBoost: the model that doesn't train won 14/14

> [Open the canonical story](<https://aidr.today/1db9646a?lang=en>)

**Published:** 2026-09-28T02:27:40.000Z
**Category:** Research
**Topics:** inference

## Summary

The claim behind TabPFN and TabICL is that they predict on a table without ever training on it and still beat tuned boosting\.

## Sources

- [Story source](<https://efraingaray.com/en/blog/tabpfn-vs-xgboost/>)
- [Discussion](<https://news.ycombinator.com/item?id=49872864>)