United States

Text written by humans makes up the vast majority of the data used to pre-train and fine-tune large language models (LLMs). Many sources of this data---like code, forum posts, personal websites, and books---are easily attributed to one or a few &quot;users&quot;. In this paper, we ask if it is possible to infer if any of a _user&#39;s_ data was used to train an LLM. Not only would this constitute a breach of privacy, but it would also enable users to detect when their data was used for training. We develop the first effective attacks for _user inference_---at times, with near-perfect success---against LLMs. Our attacks are easy to employ, requiring only black-box access to an LLM and a few samples from the user, which _need not be the ones that were trained on_. We find, both theoretically and empirically, that certain properties make users more susceptible to user inference: being an outlier, having highly correlated examples, and contributing a larger fraction of data. Based on these findings, we identify several methods for mitigating user inference including training with example-level differential privacy, removing within-user duplicate examples, and reducing a user&#39;s contribution to the training data. Though these provide partial mitigation, our work highlights the need to develop methods to fully protect LLMs from user inference.

EMNLP 2024

User Inference Attacks on Large Language Models

memory-efficient fine-tuning

full-parameter fine-tuning

language models

Text written by humans makes up the vast majority of the data used to pre-train and fine-tune large language models (LLMs). Many sources of this data---like code, forum posts, personal websites, and books---are easily attributed to one or a few "users". In this paper, we ask if it is possible to infer if any of a _user's_ data was used to train an LLM. Not only would this constitute a breach of privacy, but it would also enable users to detect when their data was used for training. We develop the first effective attacks for _user inference_---at times, with near-perfect success---against LLMs. Our attacks are easy to employ, requiring only black-box access to an LLM and a few samples from the user, which _need not be the ones that were trained on_. We find, both theoretically and empirically, that certain properties make users more susceptible to user inference: being an outlier, having highly correlated examples, and contributing a larger fraction of data. Based on these findings, we identify several methods for mitigating user inference including training with example-level differential privacy, removing within-user duplicate examples, and reducing a user's contribution to the training data. Though these provide partial mitigation, our work highlights the need to develop methods to fully protect LLMs from user inference.

technical paper

## 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.

Most NLP work on narrative detection has focused on prescriptive definitions of stories crafted by researchers, leaving open the questions: how do crowd workers perceive texts to be a story, and why? We investigate this by building StoryPerceptions, a dataset of 2,496 perceptions of storytelling in 502 social media texts from 255 crowd workers, including categorical labels along with free-text storytelling rationales, authorial intent, and more. We construct a fine-grained bottom-up taxonomy of crowd workers' varied and nuanced perceptions of storytelling by open-coding their free-text rationales. Through comparative analyses at the label and code level, we illuminate patterns of disagreement among crowd workers and across other annotation contexts, including prescriptive labeling from researchers and LLM-based predictions. Notably, plot complexity, references to generalized or abstract actions, and holistic aesthetic judgments (such as a sense of cohesion) are especially important in disagreements. Our empirical findings broaden understanding of the types, relative importance, and contentiousness of features relevant to narrative detection, highlighting opportunities for future work on reader-contextualized models of narrative reception.

The Empirical Variability of Narrative Perceptions of Social Media Texts

Investigating "Hadith" texts, crucial for theological studies and Islamic jurisprudence, presents challenges due to the linguistic complexity of Arabic, such as its complex morphology. In this paper, we propose an innovative approach to address the challenges of topic modeling in Hadith studies by utilizing the Contextualized Topic Model (CTM). Our study introduces RoBERT2VecTM, a novel neural-based approach that combines the RoBERTa transformer model with Doc2Vec, specifically targeting the semantic analysis of "Matn" (the actual content). The methodology outperforms many traditional state-of-the-art NLP models by generating more coherent and diverse Arabic topics. The diversity of the generated topics allows for further categorization, deepening the understanding of discussed concepts. Notably, our research highlights the critical impact of lemmatization and stopwords in enhancing topic modeling. This breakthrough marks a significant stride in applying NLP to non-Latin languages and opens new avenues for the nuanced analysis of complex religious texts.

RoBERT2VecTM: A Novel Approach for Topic Extraction in Islamic Studies

Existing debiasing techniques are typically training-based or require access to the model's internals and output distributions, so they are inaccessible to end-users looking to adapt LLM outputs for their particular needs. In this study, we examine whether structured prompting techniques can offer opportunities for fair text generation. We evaluate a comprehensive end-user-focused iterative framework of debiasing that applies System 2 thinking processes for prompts to induce logical, reflective, and critical text generation, with single, multi-step, instruction, and role-based variants. By systematically evaluating many LLMs across many datasets and different prompting strategies, we show that the more complex System 2-based Implicative Prompts significantly improve over other techniques demonstrating lower mean bias in the outputs with competitive performance on the downstream tasks. Our work offers research directions for the design and the potential of end-user-focused evaluative frameworks for LLM use.

Thinking Fair and Slow: On the Efficacy of Structured Prompts for Debiasing Language Models

Rotary Position Embedding (RoPE) is an efficient position encoding approach and is widely utilized in numerous large language models (LLMs). Recently, a lot of methods have been put forward to further expand the context window based on RoPE. The core concept of those methods is to predefine or search for a set of factors to rescale the base frequencies of RoPE. Nevertheless, it is quite a challenge for existing methods to predefine an optimal factor due to the exponential search space. In view of this, we introduce PSC (Phase Shift Calibration), a small module for calibrating the frequencies predefined by existing methods. With the employment of PSC, we demonstrate that many existing methods can be further enhanced, like PI, YaRN, and LongRoPE. We conducted extensive experiments across multiple models and tasks. The results demonstrate that (1) when PSC is enabled, the comparative reductions in perplexity increase as the context window size is varied from 16k, to 32k, and up to 64k. (2) Our approach is broadly applicable and exhibits robustness across a variety of models and tasks.

PSC: Extending Context Window of Large Language Models via Phase Shift Calibration

The stock market provides a rich well of information that can be split across modalities, making it an ideal candidate for multimodal evaluation. Multimodal data plays an increasingly important role in the development of machine learning and has shown to positively impact performance. But information can do more than exist across modes--- it can exist across time. How should we attend to temporal data that consists of multiple information types? This work introduces (i) the MEANT model, a Multimodal Encoder for Antecedent information and (ii) a new dataset called TempStock, which consists of price, Tweets, and graphical data with over a million Tweets from all of the companies in the S\&P 500 Index. We find that MEANT improves performance on existing baselines by over 15\%, and that the textual information affects performance far more than the visual information on our time-dependent task from our ablation study. The code and dataset will be made available upon publication.

MEANT: Multimodal Encoder for Antecedent Information

Synthetic data is widely used in speech recognition due to the availability of text-to-speech models, which facilitate adapting models to previously unseen text domains. However, existing methods suffer in performance when they fine-tune an automatic speech recognition (ASR) model on synthetic data as they suffer from the distributional shift commonly referred to as the synthetic-to-real gap. In this paper, we find that task arithmetic is effective at mitigating this gap. 
Our proposed method, $SYN2REAL$ task vector, shows an average improvement of 10.03% improvement in word error rate over baselines on the SLURP dataset. Additionally, we show that an average of $SYN2REAL$ task vectors, when we have real speeches from multiple different domains, can further adapt the original ASR model to perform better on the target text domain.

Task Arithmetic can Mitigate Synthetic-to-Real Gap in Automatic Speech Recognition

Current approaches for detecting text reuse do not focus on recontextualization, i.e., how the new context(s) of a reused text differs from its original context(s). In this paper, we propose a novel framework called TRoTR that relies on the notion of topic relatedness for evaluating the diachronic change of context in which text is reused. TRoTR includes two NLP tasks: TRiC and TRaC. TRiC is designed to evaluate the topic relatedness between a pair of recontextualizations. TRaC is designed to evaluate the overall topic variation within a set of recontextualizations. We also provide a curated TRoTR benchmark of biblical text reuse, human-annotated with topic relatedness. The benchmark exhibits an inter-annotator agreement of .811. We evaluate multiple, established SBERT models on the TRoTR tasks and find that they exhibit greater sensitivity to textual similarity than topic relatedness. Our experiments show that fine-tuning these models can mitigate such a kind of sensitivity.

TRoTR: A Framework for Evaluating the Re-contextualization of Text Reuse

Knowledge editing technology is crucial for maintaining the accuracy and timeliness of large language models (LLMs) . However, the setting of this task overlooks a significant portion of commonsense knowledge based on free-text in the real world, characterized by broad knowledge scope, long content and non instantiation. The editing objects of previous methods (e.g., MEMIT) were single token or entity, which were not suitable for commonsense knowledge in free-text form. To address the aforementioned challenges, we conducted experiments from two perspectives: knowledge localization and knowledge editing. Firstly, we introduced Knowledge Localization for Free-Text(KLFT) method, revealing the challenges associated with the distribution of commonsense knowledge in MLP and Attention layers, as well as in decentralized distribution. Next, we propose a Dynamics-aware Editing Method(DEM), which utilizes a Dynamics-aware Module to locate the parameter positions corresponding to commonsense knowledge, and uses Knowledge Editing Module to update knowledge. The DEM method fully explores the potential of the MLP and Attention layers, and successfully edits commonsense knowledge based on free-text. The experimental results indicate that the DEM can achieve excellent editing performance.

Commonsense Knowledge Editing Based on Free-Text in LLMs

Translating lyrics for musicals presents unique challenges due to the need to ensure high translation quality while adhering to singability requirements such as length and rhyme. Existing song translation approaches often prioritize these singability constraints at the expense of translation quality, which is crucial for musicals. This paper aims to enhance translation quality while maintaining key singability features. Our method consists of three main components. First, we create a dataset to train reward models for the automatic evaluation of translation quality. Second, to enhance both singability and translation quality, we implement a two-stage training process with filtering techniques. Finally, we introduce an inference-time optimization framework for translating entire songs. Extensive experiments, including both automatic and human evaluations, demonstrate significant improvements over baseline methods and validate the effectiveness of each component in our approach.

Sing it, Narrate it: Quality Musical Lyrics Translation

Chain-of-thought (CoT), tree-of-thought (ToT), and related techniques work surprisingly well in practice for some complex reasoning tasks with Large Language Models (LLMs), but why? This work seeks the underlying reasons by conducting experimental case studies and linking the performance benefits to well-established sample and computational complexity principles in machine learning. We experimented with six reasoning tasks, ranging from grade school math, air travel planning, ..., to Blocksworld. The results suggest that (i) both CoT and ToT benefit significantly from task decomposition, which breaks a complex reasoning task into a sequence of steps with low sample complexity and explicitly outlines the reasoning structure; (ii) for computationally hard reasoning tasks, the more sophisticated tree structure of ToT outperforms the linear structure of CoT; (iii) explicitly annotating important variables is important for good performance. These findings provide useful guidelines for using LLM in solving reasoning tasks in practice.

Premium content

User Inference Attacks on Large Language Models

Downloads

Next from EMNLP 2024

The Empirical Variability of Narrative Perceptions of Social Media Texts

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES