United States

This study investigates the presence of biases in large language models (LLMs), specifically focusing on how these models process and reflect inter-state conflict structures. Previous research has often lacked the standardized datasets necessary for a thorough and consistent evaluation of biases in this context. Without such datasets, it is challenging to accurately assess the impact of these biases on critical applications. To address this gap, we developed a diverse and high-quality corpus using a four-phase process. This process included generating texts based on international conflict-related keywords, enhancing emotional diversity to capture a broad spectrum of sentiments, validating the coherence and connections between texts, and conducting final quality assurance through human reviewers who are experts in natural language processing. Our analysis, conducted using this newly developed corpus, revealed subtle but significant negative biases in LLMs, particularly towards Eastern bloc countries such as Russia and China. These biases have the potential to influence decision-making processes in fields like national security and international relations, where accurate, unbiased information is crucial. The findings underscore the importance of evaluating and mitigating these biases to ensure the reliability and fairness of LLMs when applied in sensitive areas.

EMNLP 2024

Fifty shapes of BLiMP: syntactic learning curves in language models are not uniform, but sometimes unruly

corpus development

sentiment bias

international relations

large language models

security

## Welcome to EMNLP 2024! 
We are excited to welcome you to one of the most prominent conferences in the field of Natural Language Processing. This year, EMNLP 2024 is being held in a hybrid format,
offering both virtual and in-person participation in beautiful Miami. Due to a record-breaking number of submissions, we've expanded the total number of accepted papers to accommodate more cutting-edge research from around the globe.
### [Conference Handbook](https://drive.google.com/file/d/1WPROgxjLAC96AJL7Ugy0tEnYm7dkrbHt/view?usp=sharing)

You are required to register for this event. **Please register [here](https://2024.emnlp.org/registration/).** The EMNLP 2024 event page on Underline will be open to public one week prior to the event.

Please register!

EMNLP 2024 will take place in Miami, Florida from Nov 12th to Nov 16th, 2024, at the Hyatt Regency Miami Hote and on Underline for remote participants.

As NLP models become increasingly integral to decision-making processes, the need for explainability and interpretability has become paramount. In this work, we propose a framework that achieves the aforementioned by generating semantically edited inputs, known as counterfactual interventions, which change the model prediction, thus providing a form of counterfactual explanations for the model. We frame the search for optimal counterfactual interventions as a graph assignment problem and employ a GNN to solve it, thus achieving high efficiency. We test our framework on two NLP tasks - binary sentiment classification and topic classification - and show that the generated edits are contrastive, fluent and minimal, while the whole process remains significantly faster than other state-of-the-art counterfactual editors.

Optimal and efficient text counterfactuals using Graph Neural Networks

Influence of Context on Token-Expert Assignments in Mixture-of-Experts Language Models

Routing in Sparsely-gated Language Models responds to Context

Sentence embeddings from transformer models encode much linguistic information in a fixed-length vector. We investigate whether structural information -- specifically, information about chunks and their structural and semantic properties -- can be detected in these representations. We use a dataset consisting of sentences with known chunk structure, and two linguistic intelligence datasets, whose solution relies on detecting chunks and their grammatical number, and respectively, their semantic roles. Through an approach involving indirect supervision, and through analyses of the performance on the tasks and of the internal representations built during learning, we show that information about chunks and their properties can be obtained from sentence embeddings.

Are there identifiable structural parts in the sentence embedding whole?

Large language models memorize portions of their training data verbatim. Our findings indicate that models exhibit higher memorization rates both early on and at the very end of their training, with the lowest rates occurring midway through the process. This phenomenon can be attributed to the models retaining most of the examples memorized early on, while forgetting many more examples as training progresses. Interestingly, these forgotten examples are sometimes re-memorized later on, often undergoing cycles of forgetting and re-memorization. Notably, examples memorized early in training are more likely to remain consistently retained, suggesting that they become more firmly ’crystallized’ in the model’s representation. Based on these insights, we tentatively recommend placing data that is more likely to be sensitive in the middle stages of the training process.

Learning, Forgetting, Remembering: Insights From Tracking LLM Memorization During Training

For induction heads to copy forward information successfully, heads in earlier layers must first load previous token information into every hidden state, a process Olsson et al. (2022) call "key shifting". While this information is hypothesized to exist, there have been few attempts to explicitly locate it in models. In this work, we use linear probes to identify the subspaces responsible for storing previous token information in Llama-2-7b and Llama-3-8b. We show that these subspaces are causally implicated in induction by using them to "edit" previous token information and trigger random token copying in new contexts.

Inducing Induction in Llama via Linear Probe Interventions

Whether LLMs can estimate their hallucination risk based on internal states before response generation.

LLM Internal States Reveal Hallucination Risk Faced With a Query

The surge of state-of-the-art transformer-based models has undoubtedly pushed the limits of NLP model performance, excelling in a variety of tasks. We cast the spotlight on the underexplored task of Natural Language Inference (NLI), since models trained on popular well-suited datasets are susceptible to adversarial attacks, allowing subtle input interventions to mislead the model. In this work, we validate the usage of natural language explanation as a model-agnostic defence strategy through extensive experimentation: only by fine-tuning a classifier on the explanation rather than premise-hypothesis inputs, robustness under various adversarial attacks is achieved in comparison to explanation-free baselines. Moreover, since there is no standard strategy for testing the semantic validity of the generated explanations, we research the correlation of widely used language generation metrics with human perception, in order for them to serve as a proxy towards robust NLI models. Our approach is resource-efficient and reproducible without significant computational limitations.

Enhancing adversarial robustness in Natural Language Inference using explanations

Faithfulness is a critical metric to assess the reliability of explainable AI. In NLP, current methods for faithfulness evaluation are fraught with discrepancies and biases, often failing to capture the true reasoning of models. We in- troduce Adversarial Sensitivity as a novel ap- proach to faithfulness evaluation, focusing on the explainer’s response when the model is un- der adversarial attack. Our method accounts for the faithfulness of explainers by capturing sensitivity to adversarial input changes. This work addresses significant limitations in ex- isting evaluation techniques, and furthermore, quantifies faithfulness from a crucial yet under- explored paradigm.

Faithfulness and the Notion of Adversarial Sensitivity in NLP Explanations

Language models have displayed a wide array of capabilities, but the reason for their performance remains a topic of heated debate and investigation. Do these models simply recite the observed training data, or are they able to abstract away surface statistics and learn the underlying processes from which the data was generated? To investigate this question, we explore the capabilities of a GPT model in the context of Markov Decision Processes (MDPs), where the underlying transition dynamics and policies are not directly observed. The model is trained to predict the next state or action without any initial knowledge of the MDPs or the players' policies. Despite this, we present evidence that the model develops emergent representations of the underlying parameters governing the MDPs.

Transformers Learn Hidden Dynamics when Trained to Predict Markov Decision Processes

Previous research on the predictive power (PP) of surprisal and entropy has focused on determining which language models (LMs) generate estimates with the highest PP on reading times, and examining for which populations the PP is strongest. In this study, we leverage eye movement data on texts that were generated using a range of decoding strategies with different LMs. We then extract the transition scores that reflect the models' production rather than comprehension effort. This allows us to investigate the alignment of LM language production and human language comprehension. Our findings reveal that there are differences in the strength of the alignment between reading behavior and certain LM decoding strategies and that this alignment further reflects different stages of language understanding (early, late, or global processes). Although we find lower PP of transition-based measures compared to surprisal and entropy for most decoding strategies, our results provide valuable insights into which decoding strategies impose less processing effort for readers. Our code is available via https://github.com/DiLi-Lab/LM-human-alignment.

Premium content

Fifty shapes of BLiMP: syntactic learning curves in language models are not uniform, but sometimes unruly

Downloads

Next from EMNLP 2024

Optimal and efficient text counterfactuals using Graph Neural Networks

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES