pith:AWBWG7T2
Adapting Large Language Models to a Low-Resource Agglutinative Language: A Comparative Study of LoRA and QLoRA for Bashkir
QLoRA on Mistral-7B and Phi-2 reaches near full fine-tuning perplexity for Bashkir with over 40 times fewer trainable parameters.
arxiv:2605.04948 v2 · 2026-05-06 · cs.CL
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{AWBWG7T2LTPT5NTGNFJKGBBKTG}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
QLoRA applied to Mistral-7B (3.79) and Phi-2 (3.81) achieved comparable quality with over 40 times fewer trainable parameters.
That the 71k-document Bashkir corpus is sufficiently representative for both training and reliable test-set evaluation, and that tokenizer compatibility issues do not dominate the observed performance differences across architectures.
QLoRA on 7B-scale models like Mistral achieves perplexity within 0.45 of full fine-tuning on GPT-2 medium for Bashkir while using over 40 times fewer trainable parameters, though best perplexity does not guarantee coherent monolingual generation.
Receipt and verification
| First computed | 2026-06-02T03:05:05.623385Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
0583637e7a5cdf3eb6666952a3042a99b5f508b05fe35d8ac191b6f1a665396a
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/AWBWG7T2LTPT5NTGNFJKGBBKTG \
| 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: 0583637e7a5cdf3eb6666952a3042a99b5f508b05fe35d8ac191b6f1a665396a
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "42bfc4c333f2ae086c17dbfae01adefb3b03da57166a538f8b573bd63b08a958",
"cross_cats_sorted": [],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"primary_cat": "cs.CL",
"submitted_at": "2026-05-06T14:14:35Z",
"title_canon_sha256": "35df3c614deb2de9e8d5bcdbb2c94648e245f7d69825301ca81b82508341b446"
},
"schema_version": "1.0",
"source": {
"id": "2605.04948",
"kind": "arxiv",
"version": 2
}
}