AWhat changed
Injected trainable low-rank updates while freezing pretrained Transformer weights, with measured parameter and memory savings.
Scientific scope
The record is the exact eight-author ICLR 2022 paper. Its efficiency and quality results are architecture-, rank-, task- and baseline-specific; later LoRA variants, quantization combinations and universal superiority claims remain separate.
Evidence
The linked sources identify this contribution and its scope. WRS has not independently reconstructed the scientific claims. The record is the exact eight-author ICLR 2022 paper. Its efficiency and quality results are architecture-, rank-, task- and baseline-specific; later LoRA variants, quantization combinations and universal superiority claims remain separate.
Contributors
- Edward J. Hu · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Yelong Shen · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Phillip Wallis · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Zeyuan Allen-Zhu · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Yuanzhi Li · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Shean Wang · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Lu Wang · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
- Weizhu Chen · SCIENCE: Named source author of the scoped work; individual effort shares are not measured.
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction