pith:2S6NTSNW
Learning in the Fisher Subspace: A Guided Initialization for LoRA Fine-Tuning
Using Fisher curvature from downstream data to initialize LoRA subspaces improves fine-tuning performance over weight-only methods.
arxiv:2605.01046 v3 · 2026-05-01 · cs.LG
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{2S6NTSNWQHHKBUUHZG4Z4DTZIH}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
Empirical results across diverse tasks and modalities demonstrate that data-aware initialization consistently and significantly improves downstream performance over existing approaches.
That the curvature information induced by the downstream data distribution accurately identifies parameter directions whose perturbations most influence model predictions on the target objective.
Fisher information from target data provides a better criterion than weight geometry for choosing LoRA subspaces, yielding consistent performance gains on downstream tasks.
Receipt and verification
| First computed | 2026-05-28T00:05:15.211835Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
d4bcd9c9b681cea0d287c9b99e0e7941c44117f5d574d3b87a011793317aa04e
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/2S6NTSNWQHHKBUUHZG4Z4DTZIH \
| 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: d4bcd9c9b681cea0d287c9b99e0e7941c44117f5d574d3b87a011793317aa04e
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "8f6c5a4e907da9540ed1f1c0d05b1ee17aaa2c86cde169671a699ebe5d5535c9",
"cross_cats_sorted": [],
"license": "http://creativecommons.org/licenses/by/4.0/",
"primary_cat": "cs.LG",
"submitted_at": "2026-05-01T19:20:25Z",
"title_canon_sha256": "8455f47f3b3d9eb693aaa048a1d58d4bca23a47090c0e4f76cd76b736325e538"
},
"schema_version": "1.0",
"source": {
"id": "2605.01046",
"kind": "arxiv",
"version": 3
}
}