AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

TANDA: Transfer and Adapt Pre-Trained Transformer Models for Answer Sentence Selection

Siddhant Garg, Thuy Vu, Alessandro Moschitti

AAAI 2020 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 194 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
Transfer on ASNQ, then adapt on the target dataset instead of single-step fine-tuningBERT and RoBERTa answer sentence selection models · WikiQA and TREC-QA test cleanMAP: 0.813 -> 0.893; 0.857 -> 0.912Improved
Use sequential ASNQ transfer and target adaptation instead of jointly fine-tuning on ASNQ and TREC-QABERT-Base · TREC-QA test cleanMAP: 0.898 -> 0.912Improved
Add target adaptation after ASNQ transferRoBERTa-Large-MNLI · WikiQA test cleanMAP: 0.910 -> 0.920Improved
Add target adaptation after ASNQ transferRoBERTa-Large-MNLI · TREC-QA test cleanMAP: 0.880 -> 0.943Improved
Use ASNQ rather than QNLI as the transfer dataset (ablation)BERT-Base TANDA · WikiQA test cleanMAP: 0.832 -> 0.893Improved
Use ASNQ rather than QNLI as the transfer dataset (ablation)BERT-Base TANDA · TREC-QA test cleanMAP: 0.863 -> 0.912Improved
Use only ASNQ negative label 3 rather than all negative labels (ablation)BERT-Base fine-tuned on ASNQ · WikiQA test cleanMAP: 0.884 -> 0.881No clear effect
Use only ASNQ negative label 2 rather than all negative labels (ablation)BERT-Base fine-tuned on ASNQ · WikiQA test cleanMAP: 0.884 -> 0.751Hurt
Use only ASNQ negative labels 2 and 3 rather than labels 1, 2, and 3 (ablation)BERT-Base fine-tuned on ASNQ · WikiQA test cleanMAP: 0.884 -> 0.883No clear effect
Inject 20% label noise into target training data while using TANDA rather than single-step fine-tuningBERT-Base · WikiQA test cleanMAP: 22.63% -> 2.57% dropImproved
Inject 20% label noise into target training data while using TANDA rather than single-step fine-tuningBERT-Base · TREC-QA test cleanMAP: 13.88% -> 2.30% dropImproved
Use ASNQ transfer followed by NAD adaptation rather than fine-tuning only on NADBERT and RoBERTa models · Alexa Sample 1, Sample 2, and Sample 3MAP: 0.506 -> 0.585; 0.432 -> 0.474; 0.352 -> 0.391Improved

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