paper-with-me

Papers

Not All Models Localize Linguistic Knowledge in the Same Place: A Layer-wise Probing on BERToids’ Representations

2021-11-01 · EMNLP (BlackboxNLP) 2021 11 · Mohsen Fayyaz, Ehsan Aghazadeh, Ali Modarressi, Hosein Mohebbi, Mohammad Taher Pilehvar

Most of the recent works on probing representations have focused on BERT, with the presumption that the findings might be similar to the other models. In this work, we extend the probing studies to two other models in the family, namely ELECTRA and XLNet, showing that variations in the pre-training objectives or architectural choices can result in different behaviors in encoding linguistic information in the representations. Most notably, we observe that ELECTRA tends to encode linguistic knowledge in the deeper layers, whereas XLNet instead concentrates that in the earlier layers. Also, the former model undergoes a slight change during fine-tuning, whereas the latter experiences significant adjustments. Moreover, we show that drawing conclusions based on the weight mixing evaluation strategy—which is widely used in the context of layer-wise probing—can be misleading given the norm disparity of the representations across different layers. Instead, we adopt an alternative information-theoretic probing with minimum description length, which has recently been proven to provide more reliable and informative results.

📄 PDF Abstract BibTeX

Code (0)

등록된 구현이 없습니다.

Tasks

All

Similar Papers 제목 키워드 기반

Not All Models Localize Linguistic Knowledge in the Same Place: A Layer-wise Probing on BERToids' Representations

2021-09-13 · Mohsen Fayyaz, Ehsan Aghazadeh, Ali Modarressi, Hosein Mohebbi 외

Most of the recent works on probing representations have focused on BERT, with the presumption that the findings might be similar to the other models. In this work, we extend the probing studies to two other models in th…

All

How transfer learning impacts linguistic knowledge in deep NLP models?

2021-05-31 · Findings (ACL) 2021 8 · Nadir Durrani, Hassan Sajjad, Fahim Dalvi

Transfer learning from pre-trained neural language models towards downstream tasks has been a predominant theme in NLP recently. Several researchers have shown that deep NLP models learn non-trivial amount of linguistic …

DiagnosticTransfer Learning

Discovering Salient Neurons in Deep NLP Models

2022-06-27 · Nadir Durrani, Fahim Dalvi, Hassan Sajjad

While a lot of work has been done in understanding representations learned within deep NLP models and what knowledge they capture, little attention has been paid towards individual neurons. We present a technique called …

Transfer Learning

Probing LLMs for Joint Encoding of Linguistic Categories

2023-10-28 · Giulio Starace, Konstantinos Papakostas, Rochelle Choenni, Apostolos Panagiotopoulos 외

Large Language Models (LLMs) exhibit impressive performance on a range of NLP tasks, due to the general-purpose linguistic knowledge acquired during pretraining. Existing model interpretability research (Tenney et al., 2…

POS

SymLoc: Symbolic Localization of Hallucination across HaluEval and TruthfulQA

2025-11-18 · Naveen Lamba, Sanju Tiwari, Manas Gaur arxiv

LLMs still struggle with hallucination, especially when confronted with symbolic triggers like modifiers, negation, numbers, exceptions, and named entities. Yet, we lack a clear understanding of where these symbolic hall…