pith:REGQYJ4K
Mitigating Label Shift in Tabular In-Context Learning via Test-Time Posterior Adjustment
DistPFN rescales TabPFN output probabilities at test time to counteract label shift by downweighting the training class prior.
arxiv:2605.04363 v2 · 2026-05-06 · cs.LG · cs.AI
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{REGQYJ4KZCCGLDPRXANMUJYKXF}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
We propose DistPFN, the first test-time posterior adjustment method designed for tabular foundation models... demonstrating substantial improvements for various TabPFN-based models in classification tasks under label shift, while maintaining strong performance in standard settings without label shift.
That rescaling the model's output probabilities by downweighting the training prior (and optionally applying adaptive temperature) will reliably recover a better posterior under label shift without introducing new errors or requiring knowledge of the true test prior.
DistPFN is a test-time posterior adjustment that rescales TabPFN class probabilities to reduce overfitting to the training class distribution under label shift.
Formal links
Receipt and verification
| First computed | 2026-05-26T01:03:32.218964Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
890d0c278ac884658df1b81aca270ab95be0b71f6ed3580fe9102e39491105ae
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/REGQYJ4KZCCGLDPRXANMUJYKXF \
| jq -c '.canonical_record' \
| python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 890d0c278ac884658df1b81aca270ab95be0b71f6ed3580fe9102e39491105ae
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "c109a16d5c62da25cd21e50e7a7fd4033cd71f5d25383bdce86d68e4b0815685",
"cross_cats_sorted": [
"cs.AI"
],
"license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
"primary_cat": "cs.LG",
"submitted_at": "2026-05-06T00:01:47Z",
"title_canon_sha256": "670a581c6ecc480b0819f7ebc4eeed39c1c4e3f55755e8e3de1cc0dcb5a27e13"
},
"schema_version": "1.0",
"source": {
"id": "2605.04363",
"kind": "arxiv",
"version": 2
}
}