Modeling Explicit Concerning States for Reinforcement Learning in Visual Dialogue

To encourage AI agents to conduct meaningful Visual Dialogue (VD), the use of Reinforcement Learning has been proven potential. In Reinforcement Learning, it is crucial to represent states and assign rewards based on the action-caused transitions of states. However, the state representation in previous Visual Dialogue works uses the textual information only and its transitions are implicit. In this paper, we propose Explicit Concerning States (ECS) to represent what visual contents are concerned at each round and what have been concerned throughout the Visual Dialogue. ECS is modeled from multimodal information and is represented explicitly. Based on ECS, we formulate two intuitive and interpretable rewards to encourage the Visual Dialogue agents to converse on diverse and informative visual information. Experimental results on the VisDial v1.0 dataset show our method enables the Visual Dialogue agents to generate more visual coherent, less repetitive and more visual informative dialogues compared with previous methods, according to multiple automatic metrics, human study and qualitative analysis.

Similarity-Aware Fusion Network for 3D Semantic Segmentation

In this paper, we propose a similarity-aware fusion network (SAFNet) to adaptively fuse 2D images and 3D point clouds for 3D semantic segmentation. Existing fusion-based methods achieve remarkable performances by integrating information from multiple modalities. However, they heavily rely on the correspondence between 2D pixels and 3D points by projection and can only perform the information fusion in a fixed manner, and thus their performances cannot be easily migrated to a more realistic scenario where the collected data often lack strict pair-wise features for prediction. To address this, we employ a late fusion strategy where we first learn the geometric and contextual similarities between the input and back-projected (from 2D pixels) point clouds and utilize them to guide the fusion of two modalities to further exploit complementary information. Specifically, we employ a geometric similarity module (GSM) to directly compare the spatial coordinate distributions of pair-wise 3D neighborhoods, and a contextual similarity module (CSM) to aggregate and compare spatial contextual information of corresponding central points. The two proposed modules can effectively measure how much image features can help predictions, enabling the network to adaptively adjust the contributions of two modalities to the final prediction of each point. Experimental results on the ScanNetV2 benchmark demonstrate that SAFNet significantly outperforms existing state-of-the-art fusion-based approaches across various data integrity.

Global Filter Networks for Image Classification

Recent advances in self-attention and pure multi-layer perceptrons (MLP) models for vision have shown great potential in achieving promising performance with fewer inductive biases. These models are generally based on learning interaction among spatial locations from raw data. The complexity of self-attention and MLP grows quadratically as the image size increases, which makes these models hard to scale up when high-resolution features are required. In this paper, we present the Global Filter Network (GFNet), a conceptually simple yet computationally efficient architecture, that learns long-term spatial dependencies in the frequency domain with log-linear complexity. Our architecture replaces the self-attention layer in vision transformers with three key operations: a 2D discrete Fourier transform, an element-wise multiplication between frequency-domain features and learnable global filters, and a 2D inverse Fourier transform. We exhibit favorable accuracy/complexity trade-offs of our models on both ImageNet and downstream tasks. Our results demonstrate that GFNet can be a very competitive alternative to transformer-style models and CNNs in efficiency, generalization ability and robustness. Code is available at https://github.com/raoyongming/GFNet

Rethinking the Evaluation of Neural Machine Translation

The evaluation of neural machine translation systems is usually built upon generated translation of a certain decoding method (e.g., beam search) with evaluation metrics over the generated translation (e.g., BLEU). However, this evaluation framework suffers from high search errors brought by heuristic search algorithms and is limited by its nature of evaluation over one best candidate. In this paper, we propose a novel evaluation protocol, which not only avoids the effect of search errors but provides a system-level evaluation in the perspective of model ranking. In particular, our method is based on our newly proposed exact top-$k$ decoding instead of beam search. Our approach evaluates model errors by the distance between the candidate spaces scored by the references and the model respectively. Extensive experiments on WMT'14 English-German demonstrate that bad ranking ability is connected to the well-known beam search curse, and state-of-the-art Transformer models are facing serious ranking errors. By evaluating various model architectures and techniques, we provide several interesting findings. Finally, to effectively approximate the exact search algorithm with same time cost as original beam search, we present a minimum heap augmented beam search algorithm.

Evaluating Modules in Graph Contrastive Learning

The recent emergence of contrastive learning approaches facilitates the research on graph representation learning (GRL), introducing graph contrastive learning (GCL) into the literature. These methods contrast semantically similar and dissimilar sample pairs to encode the semantics into node or graph embeddings. However, most existing works only performed model-level evaluation, and did not explore the combination space of modules for more comprehensive and systematic studies. For effective module-level evaluation, we propose a framework that decomposes GCL models into four modules: (1) a sampler to generate anchor, positive and negative data samples (nodes or graphs); (2) an encoder and a readout function to get sample embeddings; (3) a discriminator to score each sample pair (anchor-positive and anchor-negative); and (4) an estimator to define the loss function. Based on this framework, we conduct controlled experiments over a wide range of architectural designs and hyperparameter settings on node and graph classification tasks. Specifically, we manage to quantify the impact of a single module, investigate the interaction between modules, and compare the overall performance with current model architectures. Our key findings include a set of module-level guidelines for GCL, e.g., simple samplers from LINE and DeepWalk are strong and robust; an MLP encoder associated with Sum readout could achieve competitive performance on graph classification. Finally, we release our implementations and results as OpenGCL, a modularized toolkit that allows convenient reproduction, standard model and module evaluation, and easy extension.

Sequence-Level Training for Non-Autoregressive Neural Machine Translation

In recent years, Neural Machine Translation (NMT) has achieved notable results in various translation tasks. However, the word-by-word generation manner determined by the autoregressive mechanism leads to high translation latency of the NMT and restricts its low-latency applications. Non-Autoregressive Neural Machine Translation (NAT) removes the autoregressive mechanism and achieves significant decoding speedup through generating target words independently and simultaneously. Nevertheless, NAT still takes the word-level cross-entropy loss as the training objective, which is not optimal because the output of NAT cannot be properly evaluated due to the multimodality problem. In this paper, we propose using sequence-level training objectives to train NAT models, which evaluate the NAT outputs as a whole and correlates well with the real translation quality. Firstly, we propose training NAT models to optimize sequence-level evaluation metrics (e.g., BLEU) based on several novel reinforcement algorithms customized for NAT, which outperforms the conventional method by reducing the variance of gradient estimation. Secondly, we introduce a novel training objective for NAT models, which aims to minimize the Bag-of-Ngrams (BoN) difference between the model output and the reference sentence. The BoN training objective is differentiable and can be calculated efficiently without doing any approximations. Finally, we apply a three-stage training strategy to combine these two methods to train the NAT model. We validate our approach on four translation tasks (WMT14 En$\leftrightarrow$De, WMT16 En$\leftrightarrow$Ro), which shows that our approach largely outperforms NAT baselines and achieves remarkable performance on all translation tasks.

Marginal Utility Diminishes: Exploring the Minimum Knowledge for BERT Knowledge Distillation

Recently, knowledge distillation (KD) has shown great success in BERT compression. Instead of only learning from the teacher's soft label as in conventional KD, researchers find that the rich information contained in the hidden layers of BERT is conducive to the student's performance. To better exploit the hidden knowledge, a common practice is to force the student to deeply mimic the teacher's hidden states of all the tokens in a layer-wise manner. In this paper, however, we observe that although distilling the teacher's hidden state knowledge (HSK) is helpful, the performance gain (marginal utility) diminishes quickly as more HSK is distilled. To understand this effect, we conduct a series of analysis. Specifically, we divide the HSK of BERT into three dimensions, namely depth, length and width. We first investigate a variety of strategies to extract crucial knowledge for each single dimension and then jointly compress the three dimensions. In this way, we show that 1) the student's performance can be improved by extracting and distilling the crucial HSK, and 2) using a tiny fraction of HSK can achieve the same performance as extensive HSK distillation. Based on the second finding, we further propose an efficient KD paradigm to compress BERT, which does not require loading the teacher during the training of student. For two kinds of student models and computing devices, the proposed KD paradigm gives rise to training speedup of 2.7x ~ 3.4x.

GTM: A Generative Triple-Wise Model for Conversational Question Generation

Generating some appealing questions in open-domain conversations is an effective way to improve human-machine interactions and lead the topic to a broader or deeper direction. To avoid dull or deviated questions, some researchers tried to utilize answer, the "future" information, to guide question generation. However, they separate a post-question-answer (PQA) triple into two parts: post-question (PQ) and question-answer (QA) pairs, which may hurt the overall coherence. Besides, the QA relationship is modeled as a one-to-one mapping that is not reasonable in open-domain conversations. To tackle these problems, we propose a generative triple-wise model with hierarchical variations for open-domain conversational question generation (CQG). Latent variables in three hierarchies are used to represent the shared background of a triple and one-to-many semantic mappings in both PQ and QA pairs. Experimental results on a large-scale CQG dataset show that our method significantly improves the quality of questions in terms of fluency, coherence and diversity over competitive baselines.

Addressing Inquiries about History: An Efficient and Practical Framework for Evaluating Open-domain Chatbot Consistency

A good open-domain chatbot should avoid presenting contradictory responses about facts or opinions in a conversational session, known as its consistency capacity. However, evaluating the consistency capacity of a chatbot is still challenging. Employing human judges to interact with chatbots on purpose to check their capacities is costly and low-efficient, and difficult to get rid of subjective bias. In this paper, we propose the Addressing Inquiries about History (AIH), an efficient and practical framework for the consistency evaluation. At the conversation stage, AIH attempts to address appropriate inquiries about the dialogue history to induce the chatbot to redeclare the historical facts or opinions. We carry out the conversation between chatbots, which is more efficient than the human-bot interaction and can also alleviate the subjective bias. In this way, we manage to rapidly obtain a dialog session that contains responses with high contradiction possibilities. At the contradiction recognition stage, we can either employ human judges or a natural language inference (NLI) model to recognize whether the answers to the inquiries are contradictory with history. Finally, we are able to rank chatbots according to the contradiction statistics. Experiments on open-domain chatbots show that our approach can efficiently and reliably assess the consistency capacity of chatbots and achieve a high ranking correlation with the human evaluation. We release the framework and hope to help improve the consistency capacity of chatbots. \footnote{\url{https://github.com/ictnlp/AIH}}

Conversations Are Not Flat: Modeling the Dynamic Information Flow across Dialogue Utterances

Nowadays, open-domain dialogue models can generate acceptable responses according to the historical context based on the large-scale pre-trained language models. However, they generally concatenate the dialogue history directly as the model input to predict the response, which we named as the flat pattern and ignores the dynamic information flow across dialogue utterances. In this work, we propose the DialoFlow model, in which we introduce a dynamic flow mechanism to model the context flow, and design three training objectives to capture the information dynamics across dialogue utterances by addressing the semantic influence brought about by each utterance in large-scale pre-training. Experiments on the multi-reference Reddit Dataset and DailyDialog Dataset demonstrate that our DialoFlow significantly outperforms the DialoGPT on the dialogue generation task. Besides, we propose the Flow score, an effective automatic metric for evaluating interactive human-bot conversation quality based on the pre-trained DialoFlow, which presents high chatbot-level correlation ($r=0.9$) with human ratings among 11 chatbots. Code and pre-trained models will be public. \footnote{\url{https://github.com/ictnlp/DialoFlow}}