Singapore

EMNLP 2023

Thorny Roses: Investigating the Dual Use Dilemma in Natural Language Processing

dual use

harms

ai ethics

checklist

survey

workshop paper

**Welcome to EMNLP 2023!**

On behalf of the EMNLP 2023 Organizing Committee, I extend a warm and heartfelt welcome to all of you to the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP). It is with immense pleasure and excitement that we gather here in Singapore, a vibrant hub of innovation and technological advancement.

The conference program is packed with insightful presentations, thought-provoking workshops, and engaging networking opportunities. In addition to the technical sessions, we have also planned several social events that will provide you with the opportunity to connect with your colleagues.

The conference is held in person at the Resorts World Convention Centre in Singapore, and available on-line with the help of Underline.

For those in Singapore, I hope that you will find time to explore the exciting Sentosa Island, Gardens by the Bay, Singapore Botanic Gardens and the many other attractions unique to Singapore.

Of course, we are grateful to our sponsors and partners for their generous support of EMNLP 2023, whose contributions make it possible for us to host this world-class event.

Yuji Matsumoto (RIKEN AIP) 
EMNLP 2023 General Chair

To access the **EMNLP 2023** event page on Underline, you need to register for the Conference. 
Please follow **[this link](https://2023.emnlp.org/registration/)** for more details.

EMNLP 2023 took place in Singapore from Dec 6th to Dec 10th, 2023.

With their increasing size, Large language models (LLMs) are becoming increasingly good at language understanding tasks. But even with high performance on specific downstream task, LLMs fail at simple linguistic tests for negation or quantifier understanding. Previous work on testing capability of LLMs on understanding quantifiers suggest that as the size of the models increase, they get better at understanding most-type quantifiers but get increasingly worse at understanding few-type quantifiers, thus presenting a case of an inverse-scaling law. In this paper, we question the claims of inverse scaling of few-type quantifier understanding in LLMs and show that it is a result of inappropriate testing methodology. We also present alternate methods to measure quantifier comprehension in LLMs and show that as the size of the models increase, these behaviours are different from what is shown in previous research. LLMs are consistently able to understand the difference between the meaning of few-type and most-type quantifiers, but when a quantifier is added to phrase, LLMs do not always take into account the meaning of the quantifier. We in fact see an inverse scaling law for most-type quantifiers, which is contrary to human psycho-linguistic experiments and previous work, where the model's understanding of most-type quantifier gets worse as the model size increases. We do this evaluation on models ranging from 125M-175B parameters, which suggests that LLMs do not do as well as expected with quantifiers and statistical co-occurrence of words still takes precedence over word meaning.

Probing Quantifier Comprehension in Large Language Models: Another Example of Inverse Scaling

Psycholinguistic studies suggested that our mental perception of events depends not only on the lexical items used to describe them, but also on the syntactic structure of the event description. More specifically, it has been argued that light verb constructions affect the perception of duration in event construal, such that the same event in this type of constructions is perceived by humans as taking less time (to give a kiss takes a shorter time than to kiss). We present two experiments with BERT using English stimuli from psycholinguistic studies to investigate the effects of the syntactic construction on event duration and event similarity. We show that i) the dimensions of BERT vectors encode a smaller value for duration for both punctive and durative events in count syntax, in line with human results; on the other hand, we also found that ii) BERT semantic similarity fails to capture the conceptual shift that durative events should undergo in count syntax

On Quick Kisses and How to Make Them Count: A Study on Event Construal in Light Verb Constructions with BERT

Differential Privacy (DP) has been tailored to address the unique challenges of text-to-text privatization. However, text-to-text privatization is known for degrading the performance of language models when trained on perturbed text. Employing a series of interpretation techniques on the internal representations extracted from BERT trained on perturbed pre-text, we intend to disentangle at the linguistic level the distortion induced by differential privacy. Experimental results from a representational similarity analysis indicate that the overall similarity of internal representations is substantially reduced. Using probing tasks to unpack this dissimilarity, we find evidence that text-to-text privatization affects the linguistic competence across several formalisms, encoding localized properties of words while falling short at encoding the contextual relationships between spans of words.

Disentangling the Linguistic Competence of Privacy-Preserving BERT

Natural language is an appealing medium for explaining how large language models process and store information, but evaluating the faithfulness of such explanations is challenging. To help address this, we develop two modes of evaluation for natural language explanations that claim individual neurons represent a concept in a text input. In the observational mode, we evaluate claims that a neuron $a$ activates on all and only input strings that refer to a concept picked out by the proposed explanation $E$. In the intervention mode, we construe $E$ as a claim that the neuron $a$ is a causal mediator of the concept denoted by $E$. We apply our framework to the GPT-4-generated explanations of GPT-2 XL neurons of Bills et al. (2023) and show that even the most confident explanations have high error rates and little to no causal efficacy. We close the paper by critically assessing whether natural language is a good choice for explanations and whether neurons are the best level of analysis.

Rigorously Assessing Natural Language Explanations of Neurons | VIDEO

Adaptation of pre-trained transformer-based language models (LMs) to new languages, domains, and tasks has become a widely adopted practice. A prominent approach for achieving this is training small modules, called adapters, on top of the LM layers, while freezing the LM parameters. Adapters have demonstrated impressive capabilities in zero-shot cross-lingual settings, multi-task learning, and integrating knowledge into pre-trained LMs. However, despite this vast success, our understanding of the internal adaptation process of LM predictions remains limited. In this ongoing work, we tackle this question by analyzing LM predictions adapted from one language (original) to another language (target) and concretely, we consider the sequence of hidden representations at the prediction position as an information stream being evolved through additive updates across the layers. Then, we analyze the adapters' updates to the prediction, making the following observations: (a) during inference, the model operates in the original language until reaching the very last layers, where the target language abruptly becomes pronounced, and (b) updates by early adapters are gradual, allowing to skip specific adapters without any decrease in performance, while the last adapters are critical for adaptation success. Last, we hypothesize that adapters operate in a dedicated residual stream subspace that becomes pronounced in the last layers, to be tested in our upcoming research.

The Hidden Space of Transformer Adapters

State-of-the-art NLP methods achieve human-like performance on many tasks, but make errors nevertheless. Characterizing these errors in easily interpretable terms gives insight into whether a classifier is prone to making systematic errors, but also gives a way to act and improve the classifier. We propose to discover those patterns of tokens that distinguish correct and erroneous predictions as to obtain global and interpretable descriptions for arbitrary NLP classifiers. We formulate the problem of finding a succinct and non-redundant set of such patterns in terms of the Minimum Description Length principle. Through an extensive set of experiments, we show that our method, Premise, performs well in practice. Unlike existing solutions, it recovers ground truth, even on highly imbalanced data over large vocabularies. In VQA and NER case studies, we confirm that it gives clear and actionable insight into the systematic errors made by NLP classifiers.

Understanding and Mitigating Classification Errors Through Interpretable Token Patterns

Due to language models' propensity to generate toxic or hateful responses, several techniques were developed to align model generations with users' preferences. Despite the effectiveness of such methods in improving the safety of model interactions, their impact on models' internal processes is still poorly understood. 
In this work, we apply popular detoxification approaches to several language models and quantify their impact on the resulting models' prompt dependence using feature attribution methods. We evaluate the effectiveness of counter-narrative fine-tuning and compare it with reinforcement learning-driven detoxification, observing differences in prompt reliance between the two methods despite their similar detoxification performances.

Let the Models Respond: Interpreting Language Model Detoxification Through the Lens of Prompt Dependence

In this talk, we address the limitations of existing natural language recommendation explanation systems by proposing a novel approach using knowledge graphs (KGs). Unlike previous methods relying on subjective item reviews, our KG-based approach generates fact-grounded, personalized explanations that consider user preferences and item features objectively. By leveraging collaborative filtering and user-item representations, our model consistently outperforms state-of-the-art models in natural language explainable recommendation, enhancing user understanding and trust in the system.

Knowledge-grounded Natural Language Recommendation Explanation

Usual model-agnostic explanation methods create artificial instances by slightly perturbing real instances, capturing shifts in model decisions. However, such methods rely on initial data and only provide explanations of the decision for these. To tackle these problems, we propose Therapy, the first global and model-agnostic explanation method adapted to text which requires no input dataset. Therapy generates texts following the distribution learned by a classifier through cooperative generation. Because it does not rely on initial samples, it allows to generate explanations even when data is absent (e.g., for confidentiality reasons). Moreover, conversely to existing methods that combine multiple local explanations into a global one, Therapy offers a global overview of the model behavior on the input space. Our experiments show that although using no input data to generate samples, Therapy provides insightful information about features used by the classifier that is competitive with the ones from methods relying on input samples and outperforms them when input samples are not specific to the studied model.

"Honey, Tell Me What's Wrong'', Global Explanation of Textual Discriminative Models through Cooperative Generation

Large language models (LLMs) that do not give consistent answers across contexts are problematic when used for tasks with expectations of consistency–e.g. question-answering, explanations, etc. Our work presents an evaluation benchmark for self-consistency in cases of under-specification where two or more answers can be correct. We conduct a series of behavioral experiments on the OpenAI model suite using an ambiguous integer sequence completion task. We find that average consistency ranges from 67% to 82%, far higher than would be predicted if a model’s consistency was random, and increases as model capability improves. Furthermore, we show that models tend to maintain self-consistency across a series of robustness checks, including prompting speaker changes and sequence length changes. These results suggest that self-consistency arises as an emergent capability without specifically training for it. Despite this, we find that models are uncalibrated when judging their own consistency, with models displaying both over- and under-confidence. We also propose a nonparametric test for determining from token output distribution whether a model assigns non-trivial probability to alternative answers. Using this test, we find that despite increases in self-consistency, models usually place significant weight on alternative, inconsistent answers. This distribution of probability mass provides evidence that even highly self-consistent models internally compute multiple possible responses.

Premium content

Thorny Roses: Investigating the Dual Use Dilemma in Natural Language Processing

Downloads

Next from EMNLP 2023

Probing Quantifier Comprehension in Large Language Models: Another Example of Inverse Scaling

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES