Reasoning about subjective natural language descriptions, such as opinions and preferences, is a challenging topic that largely remains unsolved to date. In particular, state-of-the-art large language models (LLMs) perform disappointingly in this task, show strong biases, and do not meet the interpretability requirements often needed in these kinds of applications. We propose a novel approach for reasoning about subjective knowledge that integrates potential and implicit meanings and explicitly models the relational nature of the information. We apply supervised graph learning, offer explanations for the model’s reasoning, and show that our model performs well across all 15 topics of OpinionQA, outperforming several prominent LLMs. Our detailed analysis further shows its unique advantages and the complementary nature it offers in comparison to LLMs.
论文检索
输入标题、作者或关键词，从 1,915 篇学术成果中精准定位
Document-level event extraction aims to extract structured event information from unstructured text. However, a single document often contains limited event information and the roles of different event arguments may be biased due to the influence of the information source.This paper addresses the limitations of traditional document-level event extraction by proposing the task of cross-document event extraction (CDEE) to integrate event information from multiple documents and provide a comprehensive perspective on events. We construct a novel cross-document event extraction dataset, namely CLES, which contains 20,059 documents and 37,688 mention-level events, where over 70% of them are cross-document. To address the task, we propose a CDEE pipeline that includes 5 steps, namely event extraction, coreference resolution, entity normalization, role normalization and entity-role resolution. Our CDEE pipeline achieves about 72% F1 in end-to-end cross-document event extraction, suggesting the challenge of this task and setting up a benchmark for future research. Our work builds a new line of information extraction research and will attract new research attention.
Understanding the deep semantics of images is essential in the era dominated by social media. However, current research works primarily on the superficial description of images, revealing a notable deficiency in the systematic investigation of the inherent deep semantics. In this work, we introduce DEEPEVAL, a comprehensive benchmark to assess Large Multimodal Models’ (LMMs) capacities of visual deep semantics. DEEPEVAL includes human-annotated dataset and three progressive subtasks: fine-grained description selection, in-depth title matching, and deep semantics understanding. Utilizing DEEPEVAL, we evaluate 9 open-source LMMs and GPT-4V(ision). Our evaluation demonstrates a substantial gap between the deep semantic comprehension capabilities of existing LMMs and humans. For example, GPT-4V is 30% behind humans in understanding deep semantics, even though it achieves human-comparable performance in image description. Further analysis reveals that LMM performance on DEEPEVAL varies according to the specific facets of deep semantics explored, indicating the fundamental challenges remaining in developing LMMs.
As natural language generation (NLG) models have become prevalent, systematically assessing the quality of machine-generated texts has become increasingly important. Recent studies introduce LLM-based evaluators that operate as reference-free metrics, demonstrating their capability to adeptly handle novel tasks. However, these models generally rely on a single-agent approach, which, we argue, introduces an inherent limit to their performance. This is because there exist biases in LLM agent’s responses, including preferences for certain text structure or content. In this work, we propose DEBATE, an NLG evaluation framework based on multi-agent scoring system augmented with a concept of Devil’s Advocate. Within the framework, one agent is instructed to criticize other agents’ arguments, potentially resolving the bias in LLM agent’s answers. DEBATE substantially outperforms the previous state-of-the-art methods in two meta-evaluation benchmarks in NLG evaluation, SummEval and TopicalChat. We also show that the extensiveness of debates among agents and the persona of an agent can influence the performance of evaluators.
Spoken Named Entity Recognition (NER) aims to extract entities from speech. The extracted entities can help voice assistants better understand user’s questions and instructions. However, current Chinese Spoken NER datasets are laboratory-controlled data that are collected by reading existing texts in quiet environments, rather than natural spoken data, and the texts used for reading are also limited in topics. These limitations obstruct the development of Spoken NER in more natural and common real-world scenarios. To address this gap, we introduce a real-world Chinese Spoken NER dataset (RWCS-NER), encompassing open-domain daily conversations and task-oriented intelligent cockpit instructions. We compare several mainstream pipeline approaches on RWCS-NER. The results indicate that the current methods, affected by Automatic Speech Recognition (ASR) errors, do not perform satisfactorily in real settings. Aiming to enhance Spoken NER in real-world scenarios, we propose two approaches: self-training-asr and mapping then distilling (MDistilling). Experiments show that both approaches can achieve significant improvements, particularly MDistilling. Even compared with GPT4.0, MDistilling still reaches better results. We believe that our work will advance the field of Spoken NER in real-world settings.
A summary structure is inherent to certain types of texts according to the Genre Theory of Linguistics. Such structures aid readers in efficiently locating information within summaries. However, most existing automatic summarization methods overlook the importance of summary structure, resulting in summaries that emphasize the most prominent information while omitting essential details from other sections. While a few summarizers recognize the importance of summary structure, they rely heavily on the predefined labels of summary structures in the source document and ground truth summaries. To address these shortcomings, we developed a Structured Knowledge-Guided Summarization (SKGSum) and its variant, SKGSum-W, which do not require structure labels. Instead, these methods rely on a set of automatically extracted summary points to generate summaries. We evaluate the proposed methods using three real-world datasets. The results indicate that our methods not only improve the quality of summaries, in terms of ROUGE and BERTScore, but also broaden the types of documents that can be effectively summarized.
Towards Understanding Task-agnostic Debiasing Through the Lenses of Intrinsic Bias and Forgetfulness
PDF ↗While task-agnostic debiasing provides notable generalizability and reduced reliance on downstream data, its impact on language modeling ability and the risk of relearning social biases from downstream task-specific data remain as the two most significant challenges when debiasing Pretrained Language Models (PLMs). The impact on language modeling ability can be alleviated given a high-quality and long-contextualized debiasing corpus, but there remains a deficiency in understanding the specifics of relearning biases. We empirically ascertain that the effectiveness of task-agnostic debiasing hinges on the quantitative bias level of both the task-specific data used for downstream applications and the debiased model. We empirically show that the lower bound of the bias level of the downstream fine-tuned model can be approximated by the bias level of the debiased model, in most practical cases. To gain more in-depth understanding about how the parameters of PLMs change during fine-tuning due to the forgetting issue of PLMs, we propose a novel framework which can Propagate Socially-fair Debiasing to Downstream Fine-tuning, ProSocialTuning. Our proposed framework can push the fine-tuned model to approach the bias lower bound during downstream fine-tuning, indicating that the ineffectiveness of debiasing can be alleviated by overcoming the forgetting issue through regularizing successfully debiased attention heads based on the PLMs’ bias levels from stages of pretraining and debiasing.
Chain of Thought (CoT) is significant in improving the reasoning abilities of large language models (LLMs). However, the correlation between the effectiveness of CoT and the length of reasoning steps in prompts remains largely unknown. To shed light on this, we have conducted several empirical experiments to explore the relations. Specifically, we design experiments that expand and compress the rationale reasoning steps within CoT demonstrations, while keeping all other factors constant. We have the following key findings. First, the results indicate that lengthening the reasoning steps in prompts, even without adding new information into the prompt, considerably enhances LLMs’ reasoning abilities across multiple datasets. Alternatively, shortening the reasoning steps, even while preserving the key information, significantly diminishes the reasoning abilities of models. This finding highlights the importance of the number of steps in CoT prompts and provides practical guidance to make better use of LLMs’ potential in complex problem-solving scenarios. Second, we also investigated the relationship between the performance of CoT and the rationales used in demonstrations. Surprisingly, the result shows that even incorrect rationales can yield favorable outcomes if they maintain the requisite length of inference. Third, we observed that the advantages of increasing reasoning steps are task-dependent: simpler tasks require fewer steps, whereas complex tasks gain significantly from longer inference sequences.
The rapid advancement of Large Language Models (LLMs) has demonstrated their vast potential across various domains, attributed to their extensive pretraining knowledge and exceptional generalizability. However, LLMs often encounter challenges in generating harmful content when faced with problematic prompts. To address this problem, existing work attempted to implement a gradient ascent based approach to prevent LLMs from producing harmful output. While these methods can be effective, they frequently impact the model utility in responding to normal prompts. To address this gap, we introduce Selective Knowledge negation Unlearning (SKU), a novel unlearning framework for LLMs, designed to eliminate harmful knowledge while preserving utility on normal prompts. Specifically, SKU is consisted of two stages: harmful knowledge acquisition stage and knowledge negation stage. The first stage aims to identify and acquire harmful knowledge within the model, whereas the second is dedicated to remove this knowledge. SKU selectively isolates and removes harmful knowledge in model parameters, ensuring the model’s performance remains robust on normal prompts. Our experiments conducted across various LLM architectures demonstrate that SKU identifies a good balance point between removing harmful information and preserving utility.
Cyclical Contrastive Learning Based on Geodesic for Zero-shot Cross-lingual Spoken Language Understanding
PDF ↗Owing to the scarcity of labeled training data, Spoken Language Understanding (SLU) is still a challenging task in low-resource languages. Therefore, zero-shot cross-lingual SLU attracts more and more attention. Contrastive learning is widely applied to explicitly align representations of similar sentences across different languages. However, the vanilla contrastive learning method may face two problems in zero-shot cross-lingual SLU: (1) the consistency between different languages is neglected; (2) each utterance has two different kinds of SLU labels, i.e. slot and intent, the utterances with one different label are also pushed away without any discrimination, which limits the performance. In this paper, we propose Cyclical Contrastive Learning based on Geodesic (CCLG), which introduces cyclical contrastive learning to achieve the consistency between different languages and leverages geodesic to measure the similarity to construct the positive pairs and negative pairs. Experimental results demonstrate that our proposed framework achieves the new state-of-the-art performance on MultiATIS++ and MTOP datasets, and the model analysis further verifies that CCLG can effectively transfer knowledge between different languages.
Large Vision-Language Models (LVLMs) suffer from hallucination issues, wherein the models generate plausible-sounding but factually incorrect outputs, undermining their reliability. A comprehensive quantitative evaluation is necessary to identify and understand the extent of hallucinations in these models. However, existing benchmarks are often limited in scope, focusing mainly on object hallucinations. Furthermore, current evaluation methods struggle to effectively address the subtle semantic distinctions between model outputs and reference data, as well as the balance between hallucination and informativeness. To address these issues, we introduce a multi-dimensional benchmark covering objects, attributes, and relations, with challenging images selected based on associative biases. Moreover, we propose a large language model (LLM)-based two-stage evaluation framework that generalizes the popular CHAIR metric and incorporates both faithfulness and coverage into the evaluation. Experiments on 10 established LVLMs demonstrate that our evaluation metric is more comprehensive and better correlated with humans than existing work when evaluating on our challenging human-annotated benchmark dataset. Our work also highlights the critical balance between faithfulness and coverage of model outputs, and encourages future works to address hallucinations in LVLMs while keeping their outputs informative.
Traditionally, developing new language models (LMs) capable of addressing multiple tasks involves fine-tuning pre-trained LMs using a wide collection of datasets, a process that often incurs significant computational expenses. Model merging emerges as a cost-effective alternative, allowing the integration of existing models fine-tuned on different tasks into a single model that performs well across all tasks, eliminating the need for additional training. In this paper, we propose RankMean, an algorithm for merging fine-tuned LMs without requiring any downstream data. RankMean determines merging coefficients based on the relative rankings of weight change magnitudes and applies these coefficients for module-wise integration of various fine-tuned models. Our experimental results demonstrate that RankMean outperforms existing baseline methods on multiple benchmarks. The code is available at https://github.com/VITA-Group/RankMean.
We show that content on the web is often translated into many languages, and the low quality of these multi-way translations indicates they were likely created using Machine Translation (MT). Multi-way parallel, machine generated content not only dominates the translations in lower resource languages; it also constitutes a large fraction of the total web content in those languages. We also find evidence of a selection bias in the type of content which is translated into many languages, consistent with low quality English content being translated en masse into many lower resource languages, via MT. Our work raises serious concerns about training models such as multilingual large language models on both monolingual and bilingual data scraped from the web.
UNIWIZ: A Unified Large Language Model Orchestrated Wizard for Safe Knowledge Grounded Conversations
PDF ↗Large Language Models (LLMs) have made significant progress in integrating safety and knowledge alignment. However, adversarial actors can manipulate these models into generating unsafe responses, and excessive safety alignment can lead to unintended hallucinations. To address these challenges, we introduce UniWiz, a novel 2-step data orchestration framework that unifies safety and knowledge data generation. We propose a “safety-priming” method to generate synthetic safety data and overcome safety bottlenecks. We also inject relevant knowledge into conversations by retrieving factual information from curated sources. UniWiz dataset consists of 17,638 quality-controlled conversations and 10,000 augmented preference data. Pretrained models fine-tuned on UniWiz show improvements across various metrics and outperform state-of-the-art instruction-tuned models trained on much larger datasets.
Pre-trained Language Models (PLMs), like BERT, with self-supervision objectives exhibit remarkable performance and generalization across various tasks. However, they suffer in inference latency due to their large size. To address this issue, side branches are attached at intermediate layers, enabling early inference of samples without requiring them to pass through all layers. However, the challenge is to decide which layer to infer and exit each sample so that the accuracy and latency are balanced. Moreover, the distribution of the samples to be inferred may differ from that used for training necessitating cross-domain adaptation. We propose an online learning algorithm named Cross-Domain Inference in Early Exit BERT (CeeBERT) that dynamically determines early exits of samples based on the level of confidence at each exit point. CeeBERT learns optimal thresholds from domain-specific confidence observed at intermediate layers on the fly, eliminating the need for labeled data. Experimental results on five distinct datasets with BERT and ALBERT models demonstrate CeeBERT’s ability to improve latency by reducing unnecessary computations with minimal drop in performance. By adapting to the threshold values, CeeBERT can speed up the BERT/ALBERT models by 2\times - 3.1\times with minimal drop in accuracy. The anonymized source code is available at https://github.com/Div290/CeeBERT.
Numerous recent works aim to enhance the efficacy of Large Language Models (LLMs) through strategic prompting. In particular, the Optimization by PROmpting (OPRO) approach provides state-of-the-art performance by leveraging LLMs as optimizers where the optimization task is to find instructions that maximize the task accuracy. In this paper, we revisit OPRO for automated prompting with relatively small-scale LLMs, such as LLaMa-2 family and Mistral 7B. Our investigation reveals that OPRO shows limited effectiveness in small-scale LLMs, with limited inference capabilities constraining optimization ability. We suggest future automatic prompting engineering to consider both model capabilities and computational costs. Additionally, for small-scale LLMs, we recommend direct instructions that clearly outline objectives and methodologies as robust prompt baselines, ensuring efficient and effective prompt engineering in ongoing research.
This paper introduces the ColorSwap dataset, designed to assess and improve the proficiency of multimodal models in matching objects with their colors. The dataset is comprised of 2,000 unique image-caption pairs, grouped into 1,000 examples. Each example includes a caption-image pair, along with a “color-swapped” pair. We follow the Winoground schema: the two captions in an example have the same words, but the color words have been rearranged to modify different objects. The dataset was created through a novel blend of automated caption and image generation with humans in the loop. We evaluate image-text matching (ITM) and visual language models (VLMs) and find that even the latest ones are still not robust at this task. GPT-4V and LLaVA score 72% and 42% on our main VLM metric, although they may improve with more advanced prompting techniques. On the main ITM metric, contrastive models such as CLIP and SigLIP perform close to chance (at 12% and 30%, respectively), although the non-contrastive BLIP ITM model is stronger (87%). We also find that finetuning on fewer than 2,000 examples yields significant performance gains on this out-of-distribution word-order understanding task.
Recent watermarked generation algorithms inject detectable signatures during language generation to facilitate post-hoc detection. While token-level watermarks are vulnerable to paraphrase attacks, SemStamp (Hou et al., 2023) applies watermark on the semantic representation of sentences and demonstrates promising robustness. SemStamp employs locality-sensitive hashing (LSH) to partition the semantic space with arbitrary hyperplanes, which results in a suboptimal tradeoff between robustness and speed. We propose k-SemStamp, a simple yet effective enhancement of SemStamp, utilizing k-means clustering as an alternative of LSH to partition the embedding space with awareness of inherent semantic structure. Experimental results indicate that k-SemStamp saliently improves its robustness and sampling efficiency while preserving the generation quality, advancing a more effective tool for machine-generated text detection.
In the rapidly evolving field of large language models (LLMs), data augmentation (DA) has emerged as a pivotal technique for enhancing model performance by diversifying training examples without the need for additional data collection. This survey explores the transformative impact of LLMs on DA, particularly addressing the unique challenges and opportunities they present in the context of natural language processing (NLP) and beyond. From both data and learning perspectives, we examine various strategies that utilize LLMs for data augmentation, including a novel exploration of learning paradigms where LLM-generated data is used for diverse forms of further training. Additionally, this paper highlights the primary open challenges faced in this domain, ranging from controllable data augmentation to multi-modal data augmentation. This survey highlights a paradigm shift introduced by LLMs in DA, and aims to serve as a comprehensive guide for researchers and practitioners.
SPAGHETTI: Open-Domain Question Answering from Heterogeneous Data Sources with Retrieval and Semantic Parsing
PDF ↗We introduce SPAGHETTI: Semantic Parsing Augmented Generation for Hybrid English information from Text Tables and Infoboxes, a hybrid question-answering (QA) pipeline that utilizes information from heterogeneous knowledge sources, including knowledge base, text, tables, and infoboxes. Our LLM-augmented approach achieves state-of-the-art performance on the Compmix dataset, the most comprehensive heterogeneous open-domain QA dataset, with 56.5% exact match (EM) rate. More importantly, manual analysis on a sample of the dataset suggests that SPAGHETTI is more than 90% accurate, indicating that EM is no longer suitable for assessing the capabilities of QA systems today.