mlsanityData to 7 Oct 2026

Analyzing Multi-Head Attention on Trojan BERT Models

arXiv 2406.16925 ↗First posted 12 Jun 2024cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

This project investigates the behavior of multi-head attention in Transformer models, specifically focusing on the differences between benign and trojan models in the context of sentiment analysis. Trojan attacks cause models to perform normally on clean inputs but exhibit misclassifications when presented with inputs containing predefined triggers. We characterize attention head functions in trojan and benign models, identifying specific 'trojan' heads and analyzing their behavior.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 20 of its 38 references that are in the map. First posted on arXiv 12 Jun 2024.

Builds on

20 of 38 references.

References · 20

  1. 2017Attention Is All You Needcited by 59,932
  2. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,052
  3. 2019What Does BERT Look At? An Analysis of BERT's Attentioncited by 994
  4. 2019Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Prunedcited by 844
  5. 2016Universal adversarial perturbationscited by 774
  6. 2019Universal Adversarial Triggers for Attacking and Analyzing NLPcited by 709
  7. 2020BadNL: Backdoor Attacks against NLP Models with Semantic-preserving Improvementscited by 191
  8. 2020Self-Attention Attribution: Interpreting Information Interactions Inside Transformercited by 124
  9. 2020Universal Adversarial Attacks with Natural Triggers for Text Classificationcited by 50
  10. 2022A Multimodal Transformer: Fusing Clinical Notes with Structured EHR Data for Interpretable In-Hospital Mortality Predictioncited by 48
  11. 2022A Study of the Attention Abnormality in Trojaned BERTscited by 47
  12. 2022Towards Robust Ranker for Text Retrievalcited by 37
8 more
  1. 2021GAIA: A Transfer Learning System of Object Detection that Fits Your Needscited by 27
  2. 2024Task-Agnostic Detector for Insertion-Based Backdoor Attackscited by 23
  3. 2022Defending Graph Convolutional Networks against Dynamic Graph Perturbations via Bayesian Self-supervisioncited by 16
  4. 2021On the Distribution, Sparsity, and Inference-time Quantization of Attention Values in Transformerscited by 14
  5. 2023Text-to-image Editing by Image Information Removalcited by 12
  6. 2021CVLight: Decentralized Learning for Adaptive Traffic Signal Control with Connected Vehiclescited by 11
  7. 2024Optimizing Search Advertising Strategies: Integrating Reinforcement Learning with Generalized Second-Price Auctions for Enhanced Ad Ranking and Biddingcited by 9
  8. 2019Transfer Learning for Scientific Data Chain Extraction in Small Chemical Corpus with BERT-CRF Modelcited by 8

Built on by

Not cited yet.