Recently, many studies have demonstrated that exclusively incorporating OCR-derived text and spatial layouts with large language models (LLMs) can be highly effective for document understanding tasks. However, existing methods that integrate spatial layouts with text have limitations, such as producing overly long text sequences or failing to fully leverage the autoregressive traits of LLMs. In this work, we introduce Interleaving Layout andText in a Large Language Model (LayTextLLM) for document understanding. LayTextLLM projects each bounding box to a single embedding and interleaves it with text, efficiently avoiding long sequence issues while leveraging autoregressive traits of LLMs. LayTextLLM not only streamlines the interaction of layout and textual data but also shows enhanced performance in KIE and VQA. Comprehensive benchmark evaluations reveal significant improvements of LayTextLLM, with a 15.2% increase on KIE tasks and 10.7% on VQA tasks compared to previous SOTA OCR-based LLMs. All resources are available at URL masked for anonymous review.

Recent work has demonstrated the remarkable potential of Large Language Models (LLMs) in test-time scaling. By making models think before answering, they are able to achieve much higher accuracy with extra inference computation.However, in many real-world scenarios, models are used under time constraints, where an answer should be given within a certain output length. It is unclear whether and how the reasoning ability of different LLMs remain effective under strict constraints.We take a first look at this problem by conducting an in-depth empirical study. Specifically, we test 30 LLMs on common reasoning datasets under a wide range of output length budgets, and we analyze the correlation between the inference accuracy and various properties including model type, model size, prompt style, etc. We also consider the mappings between token budgets and actual on-device latency budgets.The results have demonstrated several interesting findings regarding the budget-aware LLM reasoning ability that differ from the unconstrained situation, e.g. the optimal choices of either model size or prompt style change under different budgets. These findings offer timely evaluation to this area and practical guidance for users to deploy LLMs under real-world latency constraints.

The rapid advancements in Multimodal Large Language Models (MLLMs) have significantly enhanced capabilities in Document Understanding. However, prevailing benchmarks like DocVQA and ChartQA predominantly comprise scanned or digital documents, inadequately reflecting the intricate challenges posed by diverse real-world scenarios such as variable illumination and physical distortions. This paper introduces WildDoc, the inaugural benchmark designed specifically for assessing document understanding in natural environments. WildDoc incorporates a diverse set of manually captured document images reflecting real-world conditions and leverages document sources from established benchmarks to facilitate comprehensive comparisons with digital or scanned documents. Further, to rigorously evaluate model robustness, each document is captured four times under different conditions. Evaluations of state-of-the-art MLLMs on WildDoc expose substantial performance declines and underscore the models’ inadequate robustness compared to traditional benchmarks, highlighting the unique challenges posed by real-world document understanding.

“筛选有效表达情感的声学特征对语音情感研究至关重要。对具有相同或相似声学特征的情感,声学研究中仅使用基频和时长无法有效区分。本研究扩大声学参数的种类和数量,使用三种机器学习方法,筛选出区分情感类型的多组有效声学参数,补充和完善语音情感声学研究的声学特征集。研究发现,区分不同情感所依赖的声学参数、参数数量、参数贡献都不相同,其中频谱和信噪参数发挥重要作用。本研究为语音情感声学分析的参数选择提供参考。”

The surge in Large Language Models (LLMs) has revolutionized natural language processing, but fine-tuning them for specific tasks often encounters challenges in balancing performance and preserving general instruction-following abilities. In this paper, we posit that the distribution gap between task datasets and the LLMs serves as the primary underlying cause. To address the problem, we introduce Self-Distillation Fine-Tuning (SDFT), a novel approach that bridges the distribution gap by guiding fine-tuning with a distilled dataset generated by the model itself to match its original distribution. Experimental results on the Llama-2-chat model across various benchmarks demonstrate that SDFT effectively mitigates catastrophic forgetting while achieving comparable or superior performance on downstream tasks compared to the vanilla fine-tuning. Moreover, SDFT demonstrates the potential to maintain the helpfulness and safety alignment of LLMs. Our code is available at https://github.com/sail-sg/sdft.

Structured dropout approaches, such as attention dropout and DropHead, have been investigated to regularize the multi-head attention mechanism in Transformers. In this paper, we propose a new regularization scheme based on token-level rather than structure-level to reduce overfitting. Specifically, we devise a novel Token-Level Masking (TLM) training strategy for Transformers to regularize the connections of self-attention, which consists of two masking techniques that are effective and easy to implement. The underlying idea is to manipulate the connections between tokens in the multi-head attention via masking, where the networks are forced to exploit partial neighbors’ information to produce a meaningful representation. The generality and effectiveness of TLM are thoroughly evaluated via extensive experiments on 4 diversified NLP tasks across 18 datasets, including natural language understanding benchmark GLUE, ChineseGLUE, Chinese Grammatical Error Correction, and data-to-text generation. The results indicate that TLM can consistently outperform attention dropout and DropHead, e.g., it increases by 0.5 points relative to DropHead with BERT-large on GLUE. Moreover, TLM can establish a new record on the data-to-text benchmark Rotowire (18.93 BLEU). Our code will be publicly available at https://github.com/Young1993/tlm.

Long video content understanding poses a challenging set of research questions as it involves long-distance, cross-media reasoning and knowledge awareness. In this paper, we present a new benchmark for this problem domain, targeting the task of deep movie/TV question answering (QA) beyond previous work’s focus on simple plot summary and short video moment settings. We define several baselines based on direct retrieval of relevant context for long-distance movie QA. Observing that real-world QAs may require higher-order multi-hop inferences, we further propose a novel framework, called the DeepMaven, which extracts events, entities, and relations from the rich multimedia content in long videos to pre-construct movie knowledge graphs (movieKGs), and at the time of QA inference, complements general semantics with structured knowledge for more effective information retrieval and knowledge reasoning. We also introduce our recently collected DeepMovieQA dataset, including 1,000 long-form QA pairs from 41 hours of videos, to serve as a new and useful resource for future work. Empirical results show the DeepMaven performs competitively for both the new DeepMovieQA and the pre-existing MovieQA dataset.

Modern large-scale Pre-trained Language Models (PLMs) have achieved tremendous success on a wide range of downstream tasks. However, most of the LM pre-training objectives only focus on text reconstruction, but have not sought to learn latent-level interpretable representations of sentences. In this paper, we manage to push the language models to obtain a deeper understanding of sentences by proposing a new pre-training objective, Sparse Latent Typing, which enables the model to sparsely extract sentence-level keywords with diverse latent types. Experimental results show that our model is able to learn interpretable latent type categories in a self-supervised manner without using any external knowledge. Besides, the language model pre-trained with such an objective also significantly improves Information Extraction related downstream tasks in both supervised and few-shot settings. Our code is publicly available at https://github.com/renll/SparseLT.

The joint multiple Intent Detection (ID) and Slot Filling (SF) is a significant challenge in spoken language understanding. Because the slots in an utterance may relate to multi-intents, most existing approaches focus on utilizing task-specific components to capture the relations between intents and slots. The customized networks restrict models from modeling commonalities between tasks and generalization for broader applications. To address the above issue, we propose a Unified Generative framework (UGEN) based on a prompt-based paradigm, and formulate the task as a question-answering problem. Specifically, we design 5-type templates as instructional prompts, and each template includes a question that acts as the driver to teach UGEN to grasp the paradigm, options that list the candidate intents or slots to reduce the answer search space, and the context denotes original utterance. Through the instructional prompts, UGEN is guided to understand intents, slots, and their implicit correlations. On two popular multi-intent benchmark datasets, experimental results demonstrate that UGEN achieves new SOTA performances on full-data and surpasses the baselines by a large margin on 5-shot (28.1%) and 10-shot (23%) scenarios, which verify that UGEN is robust and effective.

Lifelong language learning aims to stream learning NLP tasks while retaining knowledge of previous tasks. Previous works based on the language model and following data-free constraint approaches have explored formatting all data as “begin token (B) + context (C) + question (Q) + answer (A)” for different tasks. However, they still suffer from catastrophic forgetting and are exacerbated when the previous task’s pseudo data is insufficient for the following reasons: (1) The model has difficulty generating task-corresponding pseudo data, and (2) A is prone to error when A and C are separated by Q because the information of the C is diminished before generating A. Therefore, we propose the Ask Question First and Replay Question (AQF-RQ), including a novel data format “BQCA” and a new training task to train pseudo questions of previous tasks. Experimental results demonstrate that AQF-RQ makes it easier for the model to generate more pseudo data that match corresponding tasks, and is more robust to both sufficient and insufficient pseudo-data when the task boundary is both clear and unclear. AQF-RQ can achieve only 0.36% lower performance than multi-task learning.

In recent years, incorporating external knowledge for response generation in open-domain conversation systems has attracted great interest. To improve the relevancy of retrieved knowledge, we propose a neural entity linking (NEL) approach. Different from formal documents, such as news, conversational utterances are informal and multi-turn, which makes it more challenging to disambiguate the entities. Therefore, we present a context-aware named entity recognition model (NER) and entity resolution (ER) model to utilize dialogue context information. We conduct NEL experiments on three open-domain conversation datasets and validate that incorporating context information improves the performance of NER and ER models. The end-to-end NEL approach outperforms the baseline by 62.8% relatively in F1 metric. Furthermore, we verify that using external knowledge based on NEL benefits the neural response generation model.

In dialog systems, the Natural Language Understanding (NLU) component typically makes the interpretation decision (including domain, intent and slots) for an utterance before the mentioned entities are resolved. This may result in intent classification and slot tagging errors. In this work, we propose to leverage Entity Resolution (ER) features in NLU reranking and introduce a novel loss term based on ER signals to better learn model weights in the reranking framework. In addition, for a multi-domain dialog scenario, we propose a score distribution matching method to ensure scores generated by the NLU reranking models for different domains are properly calibrated. In offline experiments, we demonstrate our proposed approach significantly outperforms the baseline model on both single-domain and cross-domain evaluations.

Multi-hop QA requires the machine to answer complex questions through finding multiple clues and reasoning, and provide explanatory evidence to demonstrate the machine’s reasoning process. We propose Relation Extractor-Reader and Comparator (RERC), a three-stage framework based on complex question decomposition. The Relation Extractor decomposes the complex question, and then the Reader answers the sub-questions in turn, and finally the Comparator performs numerical comparison and summarizes all to get the final answer, where the entire process itself constitutes a complete reasoning evidence path. In the 2WikiMultiHopQA dataset, our RERC model has achieved the state-of-the-art performance, with a winning joint F1 score of 53.58 on the leaderboard. All indicators of our RERC are close to human performance, with only 1.95 behind the human level in F1 score of support fact. At the same time, the evidence path provided by our RERC framework has excellent readability and faithfulness.

The growing popularity of Virtual Assistants poses new challenges for Entity Resolution, the task of linking mentions in text to their referent entities in a knowledge base. Specifically, in the shopping domain, customers tend to mention the entities implicitly (e.g., “organic milk”) rather than use the entity names explicitly, leading to a large number of candidate products. Meanwhile, for the same query, different customers may expect different results. For example, with “add milk to my cart”, a customer may refer to a certain product from his/her favorite brand, while some customers may want to re-order products they regularly purchase. Moreover, new customers may lack persistent shopping history, which requires us to enrich the connections between customers through products and their attributes. To address these issues, we propose a new framework that leverages personalized features to improve the accuracy of product ranking. We first build a cross-source heterogeneous knowledge graph from customer purchase history and product knowledge graph to jointly learn customer and product embeddings. After that, we incorporate product, customer, and history representations into a neural reranking model to predict which candidate is most likely to be purchased by a specific customer. Experiment results show that our model substantially improves the accuracy of the top ranked candidates by 24.6% compared to the state-of-the-art product search model.