{"paper":{"title":"Benchmarking Parameter-Efficient Fine-Tuning of Large Language Models for Low-Resource Tajik Text Generation with the Tajik Web Corpus","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Mistral 7B with QLoRA rank 16 reaches mean perplexity 5.03 on Tajik text generation after release of the largest open Tajik web corpus.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Mullosharaf K. Arabov","submitted_at":"2026-05-05T13:28:31Z","abstract_excerpt":"We release the Tajik Web Corpus (319k docs, 1.11B chars) and benchmark generative LLMs on prompt continuation in Tajik, a low-resource Cyrillic-script language. Seventeen configurations across nine architectures are evaluated under three fine-tuning strategies: full fine-tuning, LoRA, and QLoRA (ranks 8 and 16). Because perplexity is not directly comparable across model families with different tokenizers, generation quality is assessed through perplexity interpreted within each family, complemented by qualitative analysis performed by a native Tajik speaker. Computational cost is measured via "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Best results were achieved by Mistral 7B with QLoRA (r=16): mean perplexity 5.03, standard deviation 0.03. The novelty lies in creating the largest verified Tajik corpus and the first systematic analysis of PEFT effectiveness for Tajik text generation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That a random 10,000-document subsample from the web corpus is representative of Tajik text distribution and that perplexity alone sufficiently captures generation quality without human evaluation or downstream task metrics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Releases Tajik Web Corpus (~1.11B characters) and finds Mistral 7B with QLoRA rank 16 yields lowest perplexity of 5.03 for Tajik generation while full fine-tuning on small models causes forgetting.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Mistral 7B with QLoRA rank 16 reaches mean perplexity 5.03 on Tajik text generation after release of the largest open Tajik web corpus.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6b66de2a5a6e37fe606131a1f27dbe62ef5042f87cc73de35ba40f54785bd05c"},"source":{"id":"2605.03742","kind":"arxiv","version":2},"verdict":{"id":"1f1c4ca6-dd72-4b2e-96a1-35a4f05cb29f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T16:31:03.654143Z","strongest_claim":"Best results were achieved by Mistral 7B with QLoRA (r=16): mean perplexity 5.03, standard deviation 0.03. The novelty lies in creating the largest verified Tajik corpus and the first systematic analysis of PEFT effectiveness for Tajik text generation.","one_line_summary":"Releases Tajik Web Corpus (~1.11B characters) and finds Mistral 7B with QLoRA rank 16 yields lowest perplexity of 5.03 for Tajik generation while full fine-tuning on small models causes forgetting.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That a random 10,000-document subsample from the web corpus is representative of Tajik text distribution and that perplexity alone sufficiently captures generation quality without human evaluation or downstream task metrics.","pith_extraction_headline":"Mistral 7B with QLoRA rank 16 reaches mean perplexity 5.03 on Tajik text generation after release of the largest open Tajik web corpus."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03742/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T13:35:06.129330Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T00:31:21.265357Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:03:40.914349Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"c4e28d378dded36d83236726677ca1e749897cf06c4178ab01245f8ae1efdbb3"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}