AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Attention Biasing and Context Augmentation for Zero-Shot Control of Encoder-Decoder Transformers for Natural Language Generation

Devamanyu Hazarika, Mahdi Namazifar, Dilek Hakkani-Tür

AAAI 2022 · Architectures and efficiency › Attention and transformers · approach: Language model pretraining and induction · 9 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
Bias cross-attention toward the knowledge snippetBART-base response generation · Topical-Chatinformativeness (human evaluation): 3.43 -> 3.84Improved
Bias cross-attention toward the knowledge snippetBART-base response generation · Topical-Chat rare test setROUGEk: 0.28 -> 0.49Improved
Bias cross-attention toward the knowledge snippetBART-base response generation · Topical-Chatfluency (human evaluation)No clear effect
Bias cross-attention toward dialog history instead of knowledge (ablation)BART-base response generation · Topical-Chat rare test setROUGEk: 0.28 -> 0.20Hurt
Gradually shift cross-attention bias from dialog history toward knowledge (ablation)BART-base response generation · Topical-Chatinformativeness (human evaluation): 3.43 -> 3.80Improved
Increase knowledge bias from 1 to 50 while keeping dialog bias at 1 (ablation)BART-base response generation · Topical-Chat frequent test setPPLr: 9.66 -> 12.23Hurt
Add a question control code (ablation)BART-base question generation · Topical-Chatpercentage of responses containing questions: 29.1% -> 29.3% (↑0.2%)No clear effect
Add question control code and attention bias toward it during early decodingBART-base question generation · Topical-Chatpercentage of responses containing questions: 29.1% -> 35.2% (↑6.1%)Improved
Add feedback control code and attention bias toward it during early decodingBART-base response generation · Topical-Chatfeedback responses (%): 17.7 -> 22.1Improved
Add positive-sentiment control code without attention bias (ablation)BART-base response generation · Topical-Chatp(positive|y): 0.561±0.02 -> 0.551±0.01No clear effect
Add positive-sentiment control code and attention bias toward it during early decodingBART-base response generation · Topical-Chatp(positive|y): 0.561±0.02 -> 0.619±0.03Improved

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