research
∙
06/15/2023
Linguistic Binding in Diffusion Models: Enhancing Attribute Correspondence through Attention Map Alignment
Text-conditioned image generation models often generate incorrect associ...
share
research
∙
06/06/2023
LEACE: Perfect linear concept erasure in closed form
Concept erasure aims to remove specified features from a representation....
research
∙
05/26/2023
Few-shot Fine-tuning vs. In-context Learning: A Fair Comparison and Evaluation
Few-shot fine-tuning and in-context learning are two alternative strateg...
research
∙
05/23/2023
All Roads Lead to Rome? Exploring the Invariance of Transformers' Representations
Transformer models bring propelling advances in various NLP tasks, thus ...
research
∙
05/21/2023
Retrieving Texts based on Abstract Descriptions
In this work, we aim to connect two research areas: instruction models a...
research
∙
05/04/2023
Conformal Nucleus Sampling
Language models generate text based on successively sampling the next wo...
research
∙
10/19/2022
DALLE-2 is Seeing Double: Flaws in Word-to-Concept Mapping in Text2Image Models
We study the way DALLE-2 maps symbols (words) in the prompt to their ref...
research
∙
10/18/2022
Linear Guardedness and its Implications
Previous work on concept identification in neural representations has fo...
research
∙
08/17/2022
Visual Comparison of Language Model Adaptation
Neural language models are widely used; however, their model parameters ...
research
∙
07/28/2022
Measuring Causal Effects of Data Statistics on Language Model's `Factual' Predictions
Large amounts of training data are one of the major reasons for the high...
research
∙
04/20/2022
Analyzing Gender Representation in Multilingual Models
Multilingual language models were shown to allow for nontrivial transfer...
research
∙
01/28/2022
Adversarial Concept Erasure in Kernel Space
The representation space of neural models for textual data emerges in an...
research
∙
01/28/2022
Linear Adversarial Concept Erasure
Modern neural models trained on textual data rely on pre-trained represe...
research
∙
06/18/2021
BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models
We show that with small-to-medium training data, fine-tuning only the bi...
research
∙
06/08/2021
Neural Extractive Search
Domain experts often need to extract structured information from large c...
research
∙
05/14/2021
Counterfactual Interventions Reveal the Causal Effect of Relative Clause Representations on Agreement Prediction
When language models process syntactically complex sentences, do they us...
research
∙
03/02/2021
Contrastive Explanations for Model Interpretability
Contrastive explanations clarify why an event occurred in contrast to an...
research
∙
10/16/2020
It's not Greek to mBERT: Inducing Word-Level Translations from Multilingual BERT
Recent works have demonstrated that multilingual BERT (mBERT) learns ric...
research
∙
10/12/2020
The Extraordinary Failure of Complement Coercion Crowdsourcing
Crowdsourcing has eased and scaled up the collection of linguistic annot...
research
∙
10/11/2020
Unsupervised Distillation of Syntactic Information from Contextualized Word Representations
Contextualized word representations, such as ELMo and BERT, were shown t...
research
∙
06/01/2020
When Bert Forgets How To POS: Amnesic Probing of Linguistic Properties and MLM Predictions
A growing body of work makes use of probing in order to investigate the ...
research
∙
04/16/2020
Null It Out: Guarding Protected Attributes by Iterative Nullspace Projection
The ability to control for the kinds of information encoded in neural re...
research
∙
08/07/2019
Ab Antiquo: Proto-language Reconstruction with RNNs
Historical linguists have identified regularities in the process of hist...
research
∙
03/15/2019
Studying the Inductive Biases of RNNs with Synthetic Variations of Natural Languages
How do typological properties such as word order and morphological case ...
research
∙
09/11/2018