343 citations · 489 across the 23 of their papers we have counts for
26 papers
Shared Circuits for Shared Grammar: Tracing Subject-Verb Agreement Across Languages
Isabella Gidi, Antonio Almudévar, Core Francisco Park +2
Multilingual large language models often generalize across languages, and prior work suggests that their internal mechanisms can overlap cross-lingually. It remains unclear, howeve…
Do Activation Verbalization Methods Convey Privileged Information?
Millicent Li, Alberto Mario Ceballos Arroyo, Giordano Rogers +2
Recent interpretability methods have proposed to translate LLM internal representations into natural language descriptions using a second verbalizer LLM. This is intended to illumi…
Can Interpretation Predict Behavior on Unseen Data?
Victoria R. Li, Jenny Kaufmann, Tian Qin +3
Interpretability research often predicts model responses to targeted mechanistic interventions. But can we predict responses to unseen input data? We propose and demonstrate this a…
PolyPythias: Stability and Outliers across Fifty Language Model Pre-Training Runs
Oskar van der Wal, Pietro Lesci, Max Muller-Eberstein +4
The stability of language model pre-training and its effects on downstream performance are still understudied. Prior work shows that the training process can yield significantly di…
Sometimes I am a Tree: Data Drives Unstable Hierarchical Generalization
Tian Qin, Naomi Saphra, David Alvarez-Melis
Early in training, LMs can behave like n-gram models, but eventually they often learn tree-based syntactic rules and generalize hierarchically out of distribution (OOD). We study t…
Recite, Reconstruct, Recollect: Memorization in LMs as a Multifaceted Phenomenon
USVSN Sai Prashanth, Alvin Deng, Kyle O'Brien +9
Memorization in language models is typically treated as a homogenous phenomenon, neglecting the specifics of the memorized data. We instead model memorization as the effect of a se…
Towards out-of-distribution generalization in large-scale astronomical surveys: robust networks learn similar representations
Yash Gondhalekar, Sultan Hassan, Naomi Saphra +1
The generalization of machine learning (ML) models to out-of-distribution (OOD) examples remains a key challenge in extracting information from upcoming astronomical surveys. Inter…
Attribute Diversity Determines the Systematicity Gap in VQA
Ian Berlot-Attwell, Kumar Krishna Agrawal, A. Michael Carrell +2
Although modern neural networks often generalize to new combinations of familiar concepts, the conditions that enable such compositionality have long been an open question. In this…
First Tragedy, then Parse: History Repeats Itself in the New Era of Large Language Models
Naomi Saphra, Eve Fleisig, Kyunghyun Cho +1
Many NLP researchers are experiencing an existential crisis triggered by the astonishing success of ChatGPT and other systems based on large language models (LLMs). After such a di…
TRAM: Bridging Trust Regions and Sharpness Aware Minimization
Tom Sherborne, Naomi Saphra, Pradeep Dasigi +1
Sharpness-aware minimization (SAM) reports improving domain generalization by reducing the loss surface curvature in the parameter space. However, generalization during fine-tuning…
Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in MLMs
Angelica Chen, Ravid Shwartz-Ziv, Kyunghyun Cho +2
Most interpretability research in NLP focuses on understanding the behavior and features of a fully trained model. However, certain insights into model behavior may only be accessi…
Latent State Models of Training Dynamics
Michael Y. Hu, Angelica Chen, Naomi Saphra +1
The impact of randomness on model training is poorly understood. How do differences in data order and initialization actually manifest in the model, such that some training runs ou…
Dynamic Masking Rate Schedules for MLM Pretraining
Zachary Ankner, Naomi Saphra, Davis Blalock +2
Most works on transformers trained with the Masked Language Modeling (MLM) objective use the original BERT model's fixed masking rate of 15%. We propose to instead dynamically sche…
One Venue, Two Conferences: The Separation of Chinese and American Citation Networks
Bingchen Zhao, Yuling Gu, Jessica Zosa Forde +1
At NeurIPS, American and Chinese institutions cite papers from each other's regions substantially less than they cite endogamously. We build a citation graph to quantify this divid…
State-of-the-art generalisation research in NLP: A taxonomy and review
Dieuwke Hupkes, Mario Giulianelli, Verna Dankers +17
The ability to generalise well is one of the primary desiderata of natural language processing (NLP). Yet, what 'good generalisation' entails and how it should be evaluated is not…
Benchmarking Compositionality with Formal Languages
Josef Valvoda, Naomi Saphra, Jonathan Rawski +2
Recombining known primitive concepts into larger novel combinations is a quintessentially human cognitive capability. Whether large neural models in NLP can acquire this ability wh…
Linear Connectivity Reveals Generalization Strategies
Jeevesh Juneja, Rachit Bansal, Kyunghyun Cho +2
It is widely accepted in the mode connectivity literature that when two neural networks are trained similarly on the same data, they are connected by a path through parameter space…
The MultiBERTs: BERT Reproductions for Robustness Analysis
Thibault Sellam, Steve Yadlowsky, Jason Wei +9
Experiments with pre-trained models such as BERT are often based on a single checkpoint. While the conclusions drawn apply to the artifact tested in the experiment (i.e., the parti…
A Non-Linear Structural Probe
Jennifer C. White, Tiago Pimentel, Naomi Saphra +1
Probes are models devised to investigate the encoding of knowledge -- e.g. syntactic structure -- in contextual representations. Probes are often designed for simplicity, which has…
LSTMs Compose (and Learn) Bottom-Up
Naomi Saphra, Adam Lopez
Recent work in NLP shows that LSTM language models capture hierarchical structure in language data. In contrast to existing work, we consider the \textit{learning} process that lea…
Pareto Probing: Trading Off Accuracy for Complexity
Tiago Pimentel, Naomi Saphra, Adina Williams +1
The question of how to probe contextual word representations for linguistic structure in a way that is both principled and useful has seen significant attention recently in the NLP…
Word Interdependence Exposes How LSTMs Compose Representations
Naomi Saphra, Adam Lopez
Recent work in NLP shows that LSTM language models capture compositional structure in language data. For a closer look at how these representations are composed hierarchically, we…
Sparsity Emerges Naturally in Neural Language Models
Naomi Saphra, Adam Lopez
Concerns about interpretability, computational resources, and principled inductive priors have motivated efforts to engineer sparse neural models for NLP tasks. If sparsity is impo…
DyNet: The Dynamic Neural Network Toolkit
Graham Neubig, Chris Dyer, Yoav Goldberg +22
We describe DyNet, a toolkit for implementing neural network models based on dynamic declaration of network structure. In the static declaration strategy that is used in toolkits l…