{"paper":{"title":"Beyond LoRA vs. Full Fine-Tuning: Gradient-Guided Optimizer Routing for LLM Adaptation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Gradient routing between full and LoRA tuning beats static choices","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Boxun Li, Haozhan Tang, Kevin Kuo, Virginia Smith, Xinyin Zhang, Xiuqi Zhu","submitted_at":"2026-05-08T01:38:58Z","abstract_excerpt":"Recent literature on fine-tuning Large Language Models highlights a fundamental debate. While Full Fine-Tuning (FFT) provides the representational plasticity required for high-entropy knowledge injection, Low-Rank Adaptation (LoRA) can match or surpass FFT performance because many tasks only require updates in a low-rank space and benefit from LoRA's additional regularization. Through empirical evaluation across diverse tasks (SQL, Medical QA, and Counterfactual Knowledge) and varying language models (Gemma-3-1B, Qwen2.5-1.5B, and Qwen2.5-3B), we verify both trends and demonstrate that relying"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our evaluations show that MoLF either improves on or stays within 1.5% of the better of FFT and LoRA across all settings, while MoLF-Efficient outperforms prior adaptive LoRA approaches by up to 20% on Fact and 9% on Med and SQL.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That gradient-guided routing at the optimizer level will produce stable training dynamics and that the observed performance gains on the tested models and tasks will hold for other LLMs and domains.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MoLF routes updates between full fine-tuning and LoRA at the optimizer level to match or exceed the better of either static method, with an efficient LoRA-only variant outperforming prior adaptive approaches.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Gradient routing between full and LoRA tuning beats static choices","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d95b4e9469b7d3d5d9293102394085b8ce920d6dfe1207dd1fbf3ec7e8b3fd31"},"source":{"id":"2605.07111","kind":"arxiv","version":2},"verdict":{"id":"238107a3-c9e5-4662-bfe5-45b7242416f9","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T01:06:23.765691Z","strongest_claim":"Our evaluations show that MoLF either improves on or stays within 1.5% of the better of FFT and LoRA across all settings, while MoLF-Efficient outperforms prior adaptive LoRA approaches by up to 20% on Fact and 9% on Med and SQL.","one_line_summary":"MoLF routes updates between full fine-tuning and LoRA at the optimizer level to match or exceed the better of either static method, with an efficient LoRA-only variant outperforming prior adaptive approaches.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That gradient-guided routing at the optimizer level will produce stable training dynamics and that the observed performance gains on the tested models and tasks will hold for other LLMs and domains.","pith_extraction_headline":"Gradient routing between full and LoRA tuning beats static choices"},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.07111/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T17:01:20.208246Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T12:03:16.659944Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"7edae3803484477b5980f53e42edef3702145ff1dd67ae2e12af8a04c4ce6185"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"333d7a786c5dc6498bf5f007daacde7861e685b842781fe932fe275ad1e28763"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}