AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Less-forgetting Multi-lingual Fine-tuning

Yuren Mao, Yaobo Liang, Nan Duan, Haobo Wang, Kai Wang, Lu Chen, Yunjun Gao

NeurIPS 2022 · Vision-language and multimodal › CLIP and zero-shot · approach: Language models for zero-shot reasoning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune on multiple languages with uniform weights instead of single-language fine-tuningXLM-R base · NER zero-shotF1: 63.97±0.58 -> 67.45±0.35Improved
fine-tune on multiple languages with uniform weights instead of single-language fine-tuningXLM-R base · TyDiQA zero-shotF1: 60.66±0.34 -> 62.51±0.43Improved
LF-MLF less-forgetting weights instead of uniform weightsmulti-lingual fine-tuning · NER zero-shotF1: 67.45±0.35 -> 68.50±0.21Improved
LF-MLF instead of uniform weightsmulti-lingual fine-tuning · TyDiQA zero-shotF1: 62.51±0.43 -> 63.03±0.31Improved
LF-MLF instead of uniform weightsmulti-lingual fine-tuning · XNLI zero-shotF1: 75.17±0.12 -> 75.56±0.15Improved
LF-MLF instead of PCGradmulti-lingual fine-tuning · XNLI zero-shotF1: 74.87±0.21 -> 75.56±0.15Improved
LF-MLF instead of GradVacmulti-lingual fine-tuning · NER zero-shotF1: 67.57±0.25 -> 68.50±0.21Improved
PCGrad or GradVac instead of uniform weightsmulti-lingual fine-tuning · NER fine-tuning deF1: 87.90±0.03 -> 87.13±0.05Hurt
LF-MLF instead of single-language fine-tuning on source languagesmulti-lingual fine-tuning · NER fine-tuning frF1: 89.75±0.09 -> 89.15±0.26No clear effect
LF-MLF instead of uniform weights on source languagesmulti-lingual fine-tuning · TyDiQA-GoldP fine-tuning ruF1: 72.43±0.08 -> 71.66±0.21Hurt
LF-MLF instead of single-language fine-tuningmulti-lingual fine-tuning · TyDiQA-GoldP fine-tuning enF1: 66.03±0.31 -> 69.31±0.41Improved
increase number of source languages from 2 to 10LF-MLF and Uniform-MLF · NER zero-shot (Figure 3)F1Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy