 A polar coordinate system represents syntax in large language modelsNeural Information Processing Systems (NeurIPS), 2024 |
 Probe-Me-Not: Protecting Pre-trained Encoders from Malicious ProbingNetwork and Distributed System Security Symposium (NDSS), 2024 |
 Mechanistic?BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP (BlackBoxNLP), 2024 |
 Rethinking the Construction of Effective Metrics for Understanding the
Mechanisms of Pretrained Language ModelsConference on Empirical Methods in Natural Language Processing (EMNLP), 2023 |
 Disentangling the Linguistic Competence of Privacy-Preserving BERTBlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP (BlackboxNLP), 2023 |
 Arithmetic with Language Models: from Memorization to ComputationNeural Networks (Neural Netw.), 2023 |
 The Architectural Bottleneck PrincipleConference on Empirical Methods in Natural Language Processing (EMNLP), 2022 |
 AST-Probe: Recovering abstract syntax trees from hidden representations
of pre-trained language modelsInternational Conference on Automated Software Engineering (ASE), 2022 |
 Kernelized Concept ErasureConference on Empirical Methods in Natural Language Processing (EMNLP), 2022 |
