论文检索

Hengrui Gu, Kaixiong Zhou, Yili Wang, Ruobing Wang, Xin Wang

During pre-training, the Text-to-Image (T2I) diffusion models encode factual knowledge into their parameters. These parameterized facts enable realistic image generation, but they may become obsolete over time, thereby misrepresenting the current state of the world. Knowledge editing techniques aim to update model knowledge in a targeted way. However, facing the dual challenges posed by inadequate editing datasets and unreliable evaluation criterion, the development of T2I knowledge editing encounter difficulties in effectively generalizing injected knowledge. In this work, we design a T2I knowledge editing framework by comprehensively spanning on three phases: First, we curate a dataset CAKE, comprising paraphrase and multi-object test, to enable more fine-grained assessment on knowledge generalization. Second, we propose a novel criterion, adaptive CLIP threshold, to effectively filter out false successful images under the current criterion and achieve reliable editing evaluation. Finally, we introduce MPE, a simple but effective approach for T2I knowledge editing. Instead of tuning parameters, MPE precisely recognizes and edits the outdated part of the conditioning text-prompt to accommodate the up-to-date knowledge. A straightforward implementation of MPE (Based on in-context learning) exhibits better overall performance than previous model editors. We hope these efforts can further promote faithful evaluation of T2I knowledge editing methods.

Joao Monteiro, Étienne Marcotte, Pierre-Andre Noel, Valentina Zantedeschi, David Vazquez, Nicolas Chapados, Christopher Pal, Perouz Taslakian

Prompts are often employed to condition decoder-only language model generation on reference information. Just-in-time processing of a context is inefficient due to the quadratic cost of self-attention operations, and caching is desirable. However, caching transformer states can easily require almost as much space as the model parameters. When the right context is not known in advance, caching the prompt can be challenging. This work addresses these limitations by introducing models that, inspired by the encoder-decoder architecture, use cross-attention to condition generation on reference text without the prompt. More precisely, we leverage pre-trained decoder-only models and only train a small number of added layers. We use Question-Answering (QA) as a testbed to evaluate the ability of our models to perform conditional generation and observe that they outperform prompt-based inference methods, are comparable to fine-tuned prompted LLMs, and drastically reduce the space footprint relative to standard KV caching by two orders of magnitude. Specifically, we introduced XC-Llama which converts a pre-trained Llama 2 into an encoder-decoder architecture by integrating cross-attention layers interleaved in between existing self-attention layers.

Yeo Jin Kim, Halim Acosta, Wookhee Min, Jonathan Rowe, Bradford Mott, Snigdha Chaturvedi, James Lester

Dialogue act recognition is the task of classifying conversational utterances based on their communicative intent or function. To address this problem, we propose a novel two-phase processing approach called Dual-Process Masking. This approach streamlines the task by masking less important tokens in the input, identified through retrospective analysis of their estimated contribution during training. It enhances interpretability by using the masks applied during classification learning. Dual-Process Masking significantly improves performance over strong baselines for dialogue act recognition on a collaborative problem-solving dataset and three public dialogue benchmarks.

Krithika Ramesh, Nupoor Gandhi, Pulkit Madaan, Lisa Bauer, Charith Peris, Anjalie Field

The difficulty of anonymizing text data hinders the development and deployment of NLP in high-stakes domains that involve private data, such as healthcare and social services. Poorly anonymized sensitive data cannot be easily shared with annotators or external researchers, nor can it be used to train public models. In this work, we explore the feasibility of using synthetic data generated from differentially private language models in place of real data to facilitate the development of NLP in these domains without compromising privacy. In contrast to prior work, we generate synthetic data for real high-stakes domains, and we propose and conduct use-inspired evaluations to assess data quality. Our results show that prior simplistic evaluations have failed to highlight utility, privacy, and fairness issues in the synthetic data. Overall, our work underscores the need for further improvements to synthetic data generation for it to be a viable way to enable privacy-preserving data sharing.

Yusong Wang, Dongyuan Li, Jialun Shen, Yicheng Xu, Mingkun Xu, Kotaro Funakoshi, Manabu Okumura

Multi-modal machine translation (MMT) can reduce ambiguity and semantic distortion compared with traditional machine translation (MT) by utilizing auxiliary information such as images. However, current MMT methods face two primary challenges. The first is their underperformance compared to MT methods based on pre-trained models. The second is the inadequate exploitation and integration of the image modality within the model, primarily due to a lack of triplet training data. A mainstream approach is to introduce large amounts of parallel and monolingual data to train the text model and the visual model separately. However, incorporating extensive external data can result in data imbalance, which may introduce biases during training. Additionally, the collection and cleaning of such large datasets is labor-intensive. To overcome these challenges, we introduce a novel, low-cost, large language model-based data augmentation method called LAMBDA, which can enrich the original samples and expand the dataset without requiring external images and text. We propose a fine-grained image captioning module with a noise filter to hierarchically and accurately extract unexploited information from images. Additionally, we design two specific prompts to guide the GPT-3.5 model in generating enriched texts and the corresponding translations. The enriched samples contain diverse text and strong connections between text and images, leading to significant improvements for MMT baselines, with the highest being an increase of up to 3.83 BLEU score and 3.61 METEOR score.

Marcos V Treviso, Nuno M Guerreiro, Sweta Agrawal, Ricardo Rei, José Pombal, Tania Vaz, Helena Wu, Beatriz Silva, Daan van Stigt, Andre Martins

While machine translation (MT) systems are achieving increasingly strong performance on benchmarks, they often produce translations with errors and anomalies. Understanding these errors can potentially help improve the translation quality and user experience. This paper introduces xTower, an open large language model (LLM) built on top of TowerBase designed to provide free-text explanations for translation errors in order to guide the generation of a corrected translation. The quality of the generated explanations by xTower are assessed via both intrinsic and extrinsic evaluation. We ask expert translators to evaluate the quality of the explanations across two dimensions: relatedness towards the error span being explained and helpfulness in error understanding and improving translation quality. Extrinsically, we test xTower across various experimental setups in generating translation corrections, demonstrating significant improvements in translation quality. Our findings highlight xTower’s potential towards not only producing plausible and helpful explanations of automatic translations, but also leveraging them to suggest corrected translations.

Jared Moore, Tanvi Deshpande, Diyi Yang

Large language models (LLMs) appear to bias their survey answers toward certain values. Nonetheless, some argue that LLMs are too inconsistent to simulate particular values. Are they? To answer, we first define value consistency as the similarity of answers across 1) paraphrases of one question, 2) related questions under one topic, 3) multiple-choice and open-ended use-cases of one question, and 4) multilingual translations of a question to English, Chinese, German, and Japanese. We apply these measures to a few large, open LLMs including llama-3, as well as gpt-4o, using eight thousand questions spanning more than 300 topics. Unlike prior work, we find that models are relatively consistent across paraphrases, use-cases, translations, and within a topic. Still, some inconsistencies remain. Models are more consistent on uncontroversial topics (e.g., in the U.S., “Thanksgiving”) than on controversial ones (e.g. “euthanasia”). Base models are both more consistent compared to fine-tuned models and are uniform in their consistency across topics, while fine-tuned models are more inconsistent about some topics (e.g. “euthanasia”) than others (e.g. “Women’s rights”) like our human participants.

Camille Harris, Chijioke Mgbahurike, Neha Kumar, Diyi Yang

Dialect and gender-based biases have become an area of concern in language-dependent AI systemsincluding around automatic speech recognition (ASR) which processes speech audio into text. These potential biases raise concern for discriminatory outcomes with AI systems depending on demographic- particularly gender discrimination against women, and racial discrimination against minorities with ethnic or cultural English dialects.As such we aim to evaluate the performance of ASR systems across different genders and across dialects of English. Concretely, we take a deep dive of the performance of ASR systems on men and women across four US-based English dialects: Standard American English (SAE), African American Vernacular English (AAVE), Chicano English, and Spanglish. To do this, we construct a labeled dataset of 13 hours of podcast audio, transcribed by speakers of the represented dialects. We then evaluate zero-shot performance of different automatic speech recognition models on our dataset, and further finetune models to better understand how finetuning can impact performance. Our work fills the gap of investigating possible gender disparities within underrepresented dialects.

Sungkyung Kim, Adam Lee, Junyoung Park, Andrew Chung, Jusang Oh, Jay-Yoon Lee

Recent advancements in large language models have demonstrated enhanced capabilities in visual reasoning tasks by employing additional encoders for aligning different modalities. While the Q-Former has been widely used as a general encoder for aligning several modalities including image, video, audio, and 3D with large language models, previous works on its efficient training and the analysis of its individual components have been limited. In this work, we investigate the effectiveness of parameter efficient fine-tuning (PEFT) the Q-Former using InstructBLIP with visual reasoning benchmarks ScienceQA and IconQA. We observe that applying PEFT to the Q-Former achieves comparable performance to full fine-tuning using under 2% of the trainable parameters. Additionally, we employ AdaLoRA for dynamic parameter budget reallocation to examine the relative importance of the Q-Former’s sublayers with 4 different benchmarks. Our findings reveal that the self-attention layers are noticeably more important in perceptual visual-language reasoning tasks, and relative importance of FFN layers depends on the complexity of visual-language patterns involved in tasks. The code is available at https://github.com/AttentionX/InstructBLIP_PEFT.

Mingqian Zheng, Jiaxin Pei, Lajanugen Logeswaran, Moontae Lee, David Jurgens

Prompting serves as the major way humans interact with Large Language Models (LLM). Commercial AI systems commonly define the role of the LLM in system prompts. For example, ChatGPT uses ”You are a helpful assistant” as part of its default system prompt. Despite current practices of adding personas to system prompts, it remains unclear how different personas affect a model’s performance on objective tasks. In this study, we present a systematic evaluation of personas in system prompts. We curate a list of 162 roles covering 6 types of interpersonal relationships and 8 domains of expertise. Through extensive analysis of 4 popular families of LLMs and 2,410 factual questions, we demonstrate that adding personas in system prompts does not improve model performance across a range of questions compared to the control setting where no persona is added. Nevertheless, further analysis suggests that the gender, type, and domain of the persona can all influence the resulting prediction accuracies. We further experimented with a list of persona search strategies and found that, while aggregating results from the best persona for each question significantly improves prediction accuracy, automatically identifying the best persona is challenging, with predictions often performing no better than random selection. Overall, our findings suggest that while adding a persona may lead to performance gains in certain settings, the effect of each persona can be largely random. Code and data are available at https://github.com/Jiaxin-Pei/Prompting-with-Social-Roles.

Aiqi Jiang, Nikolas Vitsakis, Tanvi Dinkar, Gavin Abercrombie, Ioannis Konstas

Gender-Based Violence (GBV) is an increasing problem online, but existing datasets fail to capture the plurality of possible annotator perspectives or ensure the representation of affected groups. We revisit two important stages in the moderation pipeline for GBV: (1) manual data labelling; and (2) automated classification. For (1), we examine two datasets to investigate the relationship between annotator identities and attitudes and the responses they give to two GBV labelling tasks. To this end, we collect demographic and attitudinal information from crowd-sourced annotators using three validated surveys from Social Psychology. We find that higher Right Wing Authoritarianism scores are associated with a higher propensity to label text as sexist, while for Social Dominance Orientation and Neosexist Attitudes, higher scores are associated with a negative tendency to do so.For (2), we conduct classification experiments using Large Language Models and five prompting strategies, including infusing prompts with annotator information. We find: (i) annotator attitudes affect the ability of classifiers to predict their labels; (ii) including attitudinal information can boost performance when we use well-structured brief annotator descriptions; and (iii) models struggle to reflect the increased complexity and imbalanced classes of the new label sets.

Vipul Kumar Rathore, Aniruddha Deb, Ankish Kumar Chandresh, Parag Singla, Mausam

Recently, very large language models (LLMs) have shown exceptional performance on several English NLP tasks with just in-context learning (ICL), but their utility in other languages is still underexplored. We investigate their effectiveness for NLP tasks in low-resource languages (LRLs), especially in the setting of zero-labelled cross-lingual transfer (0-CLT), where no labelled training data for the target language is available – however training data from one or more related medium-resource languages (MRLs) is utilized, alongside the available unlabeled test data for a target language. We introduce Self-Supervised Prompting (SSP), a novel ICL approach tailored for the 0-CLT setting. SSP is based on the key observation that LLMs output more accurate labels if in-context exemplars are from the target language (even if their labels are slightly noisy). To operationalize this, since target language training data is not available in 0-CLT, SSP operates in two stages. In Stage I, using source MRL training data, target language’s test data is noisily labeled. In Stage II, these noisy test data points are used as exemplars in ICL for further improved labelling. Additionally, our implementation of SSP uses a novel Integer Linear Programming (ILP)-based exemplar selection that balances similarity, prediction confidence (when available) and label coverage. Experiments on three tasks and eleven LRLs (from three regions) demonstrate that SSP strongly outperforms existing SOTA fine-tuned and prompting-based baselines in 0-CLT setup.

Esteban Garces Arias, Julian Rodemann, Meimingwei Li, Christian Heumann, Matthias Aßenmacher

Despite the remarkable capabilities of large language models, generating high-quality text remains a challenging task. Numerous decoding strategies—such as beam search, sampling with temperature, top‐k sampling, nucleus (top‐p) sampling, typical decoding, contrastive decoding, and contrastive search—have been proposed to address these challenges by improving coherence, diversity, and resemblance to human-generated text. In this study, we introduce Adaptive Contrastive Search (ACS), a novel decoding strategy that extends contrastive search (CS) by incorporating an adaptive degeneration penalty informed by the model’s estimated uncertainty at each generation step. ACS aims to enhance creativity and diversity while maintaining coherence to produce high-quality outputs. Extensive experiments across various model architectures, languages, and datasets demonstrate that our approach improves both creativity and coherence, underscoring its effectiveness in text-generation tasks. We release our code, datasets, and models to facilitate further research.

Pengju Gao, Tomohiro Yamasaki, Kazunori Imoto

We propose VE-KD, a novel method that balances knowledge distillation and vocabulary expansion with the aim of training efficient domain-specific language models. Compared with traditional pre-training approaches, VE-KD exhibits competitive performance in downstream tasks while reducing model size and using fewer computational resources. Additionally, VE-KD refrains from overfitting in domain adaptation. Our experiments with different biomedical domain tasks demonstrate that VE-KD performs well compared with models such as BioBERT (+1% at HoC) and PubMedBERT (+1% at PubMedQA), with about 96% less training time. Furthermore, it outperforms DistilBERT and Adapt-and-Distill, showing a significant improvement in document-level tasks. Investigation of vocabulary size and tolerance, which are hyperparameters of our method, provides insights for further model optimization. The fact that VE-KD consistently maintains its advantages, even when the corpus size is small, suggests that it is a practical approach for domain-specific language tasks and is transferrable to different domains for broader applications.

Yeo Wei Jie, Teddy Ferdinan, Przemyslaw Kazienko, Ranjan Satapathy, Erik Cambria

Large language models (LLMs) often necessitate extensive labeled datasets and training compute to achieve impressive performance across downstream tasks. This paper explores a self-training paradigm, where the LLM autonomously curates its own labels and selectively trains on unknown data samples identified through a reference-free consistency method. Empirical evaluations demonstrate significant improvements in reducing hallucination in generation across multiple subjects. Furthermore, the selective training framework mitigates catastrophic forgetting in out-of-distribution benchmarks, addressing a critical limitation in training LLMs. Our findings suggest that such an approach can substantially reduce the dependency on large labeled datasets, paving the way for more scalable and cost-effective language model training.

Debjit Paul, Robert West, Antoine Bosselut, Boi Faltings

Large language models (LLMs) have been shown to perform better when asked to reason step-by-step before answering a question. However, it is unclear to what degree the model’s final answer is faithful to the stated reasoning steps. In this paper, we perform a causal mediation analysis on twelve LLMs to examine how intermediate reasoning steps generated by the LLM influence the final outcome and find that LLMs do not reliably use their intermediate reasoning steps when generating an answer. To address this issue, we introduce FRODO, a framework to tailor small-sized LMs to generate correct reasoning steps and robustly reason over these steps. FRODO consists of an inference module that learns to generate correct reasoning steps using an implicit causal reward function and a reasoning module that learns to faithfully reason over these intermediate inferences using a counterfactual and causal preference objective. Our experiments show that FRODO significantly outperforms four competitive baselines. Furthermore, FRODO improves the robustness and generalization ability of the reasoning LM, yielding higher performance on out-of-distribution test sets. Finally, we find that FRODO’s rationales are more faithful to its final answer predictions than standard supervised fine-tuning.

Candida Maria Greco, Lucio La Cava, Andrea Tagarelli

Verbs form the backbone of language, providing the structure and meaning to sentences. Yet, their intricate semantic nuances pose a longstanding challenge. Understanding verb relations through the concept of lexical entailment is crucial for comprehending sentence meanings and grasping verb dynamics. This work investigates the capabilities of eight Large Language Models in recognizing lexical entailment relations among verbs through differently devised prompting strategies and zero-/few-shot settings over verb pairs from two lexical databases, namely WordNet and HyperLex. Our findings unveil that the models can tackle the lexical entailment recognition task with moderately good performance, although at varying degree of effectiveness and under different conditions. Also, utilizing few-shot prompting can enhance the models’ performance. However, perfectly solving the task arises as an unmet challenge for all examined LLMs, which raises an emergence for further research developments on this topic.

Marcio Monteiro, Charu Karakkaparambil James, Marius Kloft, Sophie Fellenz

Fine-tuning pretrained language models on task-specific data is a common practice in Natural Language Processing (NLP) applications. However, the number of pretrained models available to choose from can be very large, and it remains unclear how to select the optimal model without spending considerable amounts of computational resources, especially for the text domain. To address this problem, we introduce PsyMatrix, a novel framework designed to efficiently characterize text datasets. PsyMatrix evaluates multiple dimensions of text and discourse, producing interpretable, low-dimensional embeddings. Our framework has been tested using a meta-dataset repository that includes the performance of 24 pretrained large language models fine-tuned across 146 classification datasets. Using the proposed embeddings, we successfully developed a meta-learning system capable of recommending the most effective pretrained models (optimal and near-optimal) for fine-tuning on new datasets.

Byoungjip Kim, Youngsoo Jang, Lajanugen Logeswaran, Geon-Hyeong Kim, Yu Jin Kim, Honglak Lee, Moontae Lee

Large language models (LLMs) have shown the ability to solve complex decision-making tasks beyond natural language processing tasks. LLM agents based on few-shot in-context learning (ICL) achieve surprisingly high performance without training. Despite their simplicity and generalizability, ICL-based agents are limited in their ability to incorporate feedback from an environment. In this paper, we introduce Prospector, an LLM agent that consists of two complementary LLMs, an Actor and a Critic. To elicit better instruction-aligned actions from the LLM agent, we propose AskAct prompting that performs an additional self-asking step such as goal and progress checking before generating an action. Furthermore, to implicitly incorporate the environment feedback, we propose Trajectory Ranking that orders generated trajectories by predicting the expected total reward. Prospector encourages the LLM Actor to generate diverse (creative) trajectories, and harnesses the LLM Critic to select the most rewarding trajectory. On representative decision-making benchmark environments such as ALFWorld and WebShop, we empirically demonstrate that Prospector can considerably increase the success rate of given tasks, while outperforming recent advancements such as ReAct and Reflexion.

P Aditya Sreekar, Sahil Verma, Suransh Chopra, Abhishek Persad, Sarik Ghazarian, Narayanan Sadagopan

Large Language Models (LLMs) are widely used in both industry and academia for various tasks, yet evaluating the consistency of generated text responses continues to be a challenge. Traditional metrics like ROUGE and BLEU show a weak correlation with human judgment. More sophisticated metrics using Natural Language Inference (NLI) have shown improved correlations but are complex to implement, require domain-specific training due to poor cross-domain generalization, and lack explainability. More recently, prompt-based metrics using LLMs as evaluators have emerged; while they are easier to implement, they still lack explainability and depend on task-specific prompts, which limits their generalizability. This work introduces Automated eXplainable Consistency Evaluation using LLMs (AXCEL), a prompt-based consistency metric which offers explanations for the consistency scores by providing detailed reasoning and pinpointing inconsistent text spans. AXCEL is also a generalizable metric which can be adopted to multiple tasks without changing the prompt. AXCEL outperforms both non-prompt and prompt-based state-of-the-art (SOTA) metrics in detecting inconsistencies across summarization by 8.7%, free text generation by 6.2%, and data-to-text conversion tasks by 29.4%. We also evaluate the influence of underlying LLMs on prompt based metric performance and recalibrate the SOTA prompt-based metrics with the latest LLMs for fair comparison. Further, we show that AXCEL demonstrates strong performance using open source LLMs.