Tao Ge

Yadong Zhang, Shaoguang Mao, Wenshan Wu et al.

This paper introduces BI-Directional DEliberation Reasoning (BIDDER), a novel reasoning approach to enhance the decision rationality of language models. Traditional reasoning methods typically rely on historical information and employ uni-directional (left-to-right) reasoning strategy. This lack of bi-directional deliberation reasoning results in limited awareness of potential future outcomes and insufficient integration of historical context, leading to suboptimal decisions. BIDDER addresses this gap by incorporating principles of rational decision-making, specifically managing uncertainty and predicting expected utility. Our approach involves three key processes: Inferring hidden states to represent uncertain information in the decision-making process from historical data; Using these hidden states to predict future potential states and potential outcomes; Integrating historical information (past contexts) and long-term outcomes (future contexts) to inform reasoning. By leveraging bi-directional reasoning, BIDDER ensures thorough exploration of both past and future contexts, leading to more informed and rational decisions. We tested BIDDER's effectiveness in two well-defined scenarios: Poker (Limit Texas Hold'em) and Negotiation. Our experiments demonstrate that BIDDER significantly improves the decision-making capabilities of LLMs and LLM agents.

Xu Ouyang, Tao Ge, Thomas Hartvigsen et al.

We reveal that low-bit quantization favors undertrained large language models (LLMs) by observing that models with larger sizes or fewer training tokens experience less quantization-induced degradation (QiD) when applying low-bit quantization, whereas smaller models with extensive training tokens suffer significant QiD. To gain deeper insights into this trend, we study over 1500 quantized LLM checkpoints of various sizes and at different training levels (undertrained or fully trained) in a controlled setting, deriving scaling laws for understanding the relationship between QiD and factors such as the number of training tokens, model size and bit width. With the derived scaling laws, we propose a novel perspective that we can use QiD to measure an LLM's training levels and determine the number of training tokens required for fully training LLMs of various sizes. Moreover, we use the scaling laws to predict the quantization performance of different-sized LLMs trained with 100 trillion tokens. Our projection shows that the low-bit quantization performance of future models, which are expected to be trained with over 100 trillion tokens, may NOT be desirable. This poses a potential challenge for low-bit quantization in the future and highlights the need for awareness of a model's training level when evaluating low-bit quantization research. To facilitate future research on this problem, we release all the 1500+ quantized checkpoints used in this work at https://huggingface.co/Xu-Ouyang.

Shwai He, Tao Ge, Guoheng Sun et al.

Traditional transformer models often allocate a fixed amount of computational resources to every input token, leading to inefficient and unnecessary computation. To address this, the Mixture of Depths (MoD) was introduced to dynamically adjust the computational depth by skipping less important layers. Despite its promise, current MoD approaches remain under-explored and face two main challenges: (1) high training costs due to the need to train the entire model along with the routers that determine which layers to skip, and (2) the risk of performance degradation when important layers are bypassed. In response to the first issue, we propose Router-Tuning, a method that fine-tunes only the router on a small dataset, drastically reducing the computational overhead associated with full model training. For the second challenge, we propose MindSkip, which deploys Attention with Dynamic Depths. This method preserves the model's performance while significantly enhancing computational and memory efficiency. Extensive experiments demonstrate that our approach delivers competitive results while dramatically improving the computation efficiency, e.g., 21\% speedup and only a 0.2\% performance drop. The code is released at https://github.com/CASE-Lab-UMD/Router-Tuning.

Heming Xia, Zhe Yang, Qingxiu Dong et al. · pku

To mitigate the high inference latency stemming from autoregressive decoding in Large Language Models (LLMs), Speculative Decoding has emerged as a novel decoding paradigm for LLM inference. In each decoding step, this method first drafts several future tokens efficiently and then verifies them in parallel. Unlike autoregressive decoding, Speculative Decoding facilitates the simultaneous decoding of multiple tokens per step, thereby accelerating inference. This paper presents a comprehensive overview and analysis of this promising decoding paradigm. We begin by providing a formal definition and formulation of Speculative Decoding. Then, we organize in-depth discussions on its key facets, such as drafter selection and verification strategies. Furthermore, we present a comparative analysis of leading methods under third-party testing environments. We aim for this work to serve as a catalyst for further research on Speculative Decoding, ultimately contributing to more efficient LLM inference.

Chenshuo Wang, Shaoguang Mao, Tao Ge et al.

Enhancing word usage is a desired feature for writing assistance. To further advance research in this area, this paper introduces "Smart Word Suggestions" (SWS) task and benchmark. Unlike other works, SWS emphasizes end-to-end evaluation and presents a more realistic writing assistance scenario. This task involves identifying words or phrases that require improvement and providing substitution suggestions. The benchmark includes human-labeled data for testing, a large distantly supervised dataset for training, and the framework for evaluation. The test data includes 1,000 sentences written by English learners, accompanied by over 16,000 substitution suggestions annotated by 10 native speakers. The training dataset comprises over 3.7 million sentences and 12.7 million suggestions generated through rules. Our experiments with seven baselines demonstrate that SWS is a challenging task. Based on experimental analysis, we suggest potential directions for future research on SWS. The dataset and related codes is available at https://github.com/microsoft/SmartWordSuggestions.

Xin Sun, Tao Ge, Furu Wei et al.

In this paper, we propose Shallow Aggressive Decoding (SAD) to improve the online inference efficiency of the Transformer for instantaneous Grammatical Error Correction (GEC). SAD optimizes the online inference efficiency for GEC by two innovations: 1) it aggressively decodes as many tokens as possible in parallel instead of always decoding only one token in each step to improve computational parallelism; 2) it uses a shallow decoder instead of the conventional Transformer architecture with balanced encoder-decoder depth to reduce the computational cost during inference. Experiments in both English and Chinese GEC benchmarks show that aggressive decoding could yield the same predictions as greedy decoding but with a significant speedup for online inference. Its combination with the shallow decoder could offer an even higher online inference speedup over the powerful Transformer baseline without quality loss. Not only does our approach allow a single model to achieve the state-of-the-art results in English GEC benchmarks: 66.4 F0.5 in the CoNLL-14 and 72.9 F0.5 in the BEA-19 test set with an almost 10x online inference speedup over the Transformer-big model, but also it is easily adapted to other languages. Our code is available at https://github.com/AutoTemp/Shallow-Aggressive-Decoding.

Canwen Xu, Wangchunshu Zhou, Tao Ge et al.

Cant is important for understanding advertising, comedies and dog-whistle politics. However, computational research on cant is hindered by a lack of available datasets. In this paper, we propose a large and diverse Chinese dataset for creating and understanding cant from a computational linguistics perspective. We formulate a task for cant understanding and provide both quantitative and qualitative analysis for tested word embedding similarity and pretrained language models. Experiments suggest that such a task requires deep language understanding, common sense, and world knowledge and thus can be a good testbed for pretrained language models and help models perform better on other tasks. The code is available at https://github.com/JetRunner/dogwhistle. The data and leaderboard are available at https://competitions.codalab.org/competitions/30451.

Yadong Zhang, Shaoguang Mao, Tao Ge et al.

Strategic reasoning is a complex yet essential capability for intelligent agents. It requires Large Language Model (LLM) agents to adapt their strategies dynamically in multi-agent environments. Unlike static reasoning tasks, success in these contexts depends on anticipating other agents' beliefs and actions while continuously adjusting strategies to achieve individual goals. LLMs and LLM agents often struggle with strategic reasoning due to the absence of a reasoning framework that enables them to dynamically infer others' perspectives and adapt to changing environments. Inspired by the Level-K framework from game theory and behavioral economics, which extends reasoning from simple reactions to structured strategic depth, we propose a novel framework: "K-Level Reasoning with Large Language Models (K-R)." This framework employs recursive mechanisms to enable LLMs to achieve varying levels of strategic depth, allowing agents to form higher order beliefs - beliefs about others' beliefs. We validate this framework through rigorous testing on four testbeds: two classical game theory problems and two social intelligence tasks. The results demonstrate the advantages of K-R in strategic reasoning. Our work presents the first recursive implementation of strategic depth in large language models (LLMs). It establishes a foundation for future research into theory of mind and strategic reasoning in LLMs.

Xiaoyang Wang, Hongming Zhang, Tao Ge et al.

Customizable role-playing in large language models (LLMs), also known as character generalization, is gaining increasing attention for its versatility and cost-efficiency in developing and deploying role-playing dialogue agents. This study explores a large-scale data synthesis approach to equip LLMs with character generalization capabilities. We begin by synthesizing large-scale character profiles using personas from Persona Hub and then explore two strategies: response rewriting and response generation, to create character-aligned instructional responses. To validate the effectiveness of our synthetic instruction tuning data for character generalization, we perform supervised fine-tuning (SFT) using the LLaMA-3 8B model. Our best-performing model strengthens the original LLaMA-3 8B Instruct model and achieves performance comparable to GPT-4o models on role-playing dialogue. We release our synthetic characters and instruction-tuning dialogues to support public research.

Canwen Xu, Wangchunshu Zhou, Tao Ge et al.

Recent studies on compression of pretrained language models (e.g., BERT) usually use preserved accuracy as the metric for evaluation. In this paper, we propose two new metrics, label loyalty and probability loyalty that measure how closely a compressed model (i.e., student) mimics the original model (i.e., teacher). We also explore the effect of compression with regard to robustness under adversarial attacks. We benchmark quantization, pruning, knowledge distillation and progressive module replacing with loyalty and robustness. By combining multiple compression techniques, we provide a practical strategy to achieve better accuracy, loyalty and robustness.

Tao Ge, Maria Medrano, Rui Liao et al.

Dual-energy CT (DECT) has been widely investigated to generate more informative and more accurate images in the past decades. For example, Dual-Energy Alternating Minimization (DEAM) algorithm achieves sub-percentage uncertainty in estimating proton stopping-power mappings from experimental 3-mm collimated phantom data. However, elapsed time of iterative DECT algorithms is not clinically acceptable, due to their low convergence rate and the tremendous geometry of modern helical CT scanners. A CNN-based initialization method is introduced to reduce the computational time of iterative DECT algorithms. DEAM is used as an example of iterative DECT algorithms in this work. The simulation results show that our method generates denoised images with greatly improved estimation accuracy for adipose, tonsils, and muscle tissue. Also, it reduces elapsed time by approximately 5-fold for DEAM to reach the same objective function value for both simulated and real data.

Wangchunshu Zhou, Tao Ge, Canwen Xu et al.

In this paper, we generalize text infilling (e.g., masked language models) by proposing Sequence Span Rewriting (SSR) as a self-supervised sequence-to-sequence (seq2seq) pre-training objective. SSR provides more fine-grained learning signals for text representations by supervising the model to rewrite imperfect spans to ground truth, and it is more consistent than text infilling with many downstream seq2seq tasks that rewrite a source sentences into a target sentence. Our experiments with T5 models on various seq2seq tasks show that SSR can substantially improve seq2seq pre-training. Moreover, we observe SSR is especially helpful to improve pre-training a small-size seq2seq model with a powerful imperfect span generator, which indicates a new perspective of transferring knowledge from a large model to a smaller model for seq2seq pre-training.

Mengyun Chen, Tao Ge, Xingxing Zhang et al.

We propose a novel language-independent approach to improve the efficiency for Grammatical Error Correction (GEC) by dividing the task into two subtasks: Erroneous Span Detection (ESD) and Erroneous Span Correction (ESC). ESD identifies grammatically incorrect text spans with an efficient sequence tagging model. Then, ESC leverages a seq2seq model to take the sentence with annotated erroneous spans as input and only outputs the corrected text for these spans. Experiments show our approach performs comparably to conventional seq2seq approaches in both English and Chinese GEC benchmarks with less than 50% time cost for inference.

Wangchunshu Zhou, Canwen Xu, Tao Ge et al.

In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers remain unchanged for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimental results with an ALBERT model show that our method can improve the accuracy and robustness of the model by preventing it from overthinking and exploiting multiple classifiers for prediction, yielding a better accuracy-speed trade-off compared to existing early exit methods.

Yi Zhang, Tao Ge, Xu Sun

The main barrier to progress in the task of Formality Style Transfer is the inadequacy of training data. In this paper, we study how to augment parallel data and propose novel and simple data augmentation methods for this task to obtain useful sentence pairs with easily accessible models and systems. Experiments demonstrate that our augmented parallel data largely helps improve formality style transfer when it is used to pre-train the model, leading to the state-of-the-art results in the GYAFC benchmark dataset.

Wangchunshu Zhou, Tao Ge, Ke Xu et al.

In this paper, we introduce DropHead, a structured dropout method specifically designed for regularizing the multi-head attention mechanism, which is a key component of transformer, a state-of-the-art model for various NLP tasks. In contrast to the conventional dropout mechanisms which randomly drop units or connections, the proposed DropHead is a structured dropout method. It drops entire attention-heads during training and It prevents the multi-head attention model from being dominated by a small portion of attention heads while also reduces the risk of overfitting the training data, thus making use of the multi-head attention mechanism more efficiently. Motivated by recent studies about the learning dynamic of the multi-head attention mechanism, we propose a specific dropout rate schedule to adaptively adjust the dropout rate of DropHead and achieve better regularization effect. Experimental results on both machine translation and text classification benchmark datasets demonstrate the effectiveness of the proposed approach.

Canwen Xu, Wangchunshu Zhou, Tao Ge et al.

In this paper, we propose a novel model compression approach to effectively compress BERT by progressive module replacing. Our approach first divides the original BERT into several modules and builds their compact substitutes. Then, we randomly replace the original modules with their substitutes to train the compact modules to mimic the behavior of the original modules. We progressively increase the probability of replacement through the training. In this way, our approach brings a deeper level of interaction between the original and compact models. Compared to the previous knowledge distillation approaches for BERT compression, our approach does not introduce any additional loss function. Our approach outperforms existing knowledge distillation approaches on GLUE benchmark, showing a new perspective of model compression.

Wangchunshu Zhou, Tao Ge, Ke Xu

Local sequence transduction (LST) tasks are sequence transduction tasks where there exists massive overlapping between the source and target sequences, such as Grammatical Error Correction (GEC) and spell or OCR correction. Previous work generally tackles LST tasks with standard sequence-to-sequence (seq2seq) models that generate output tokens from left to right and suffer from the issue of unbalanced outputs. Motivated by the characteristic of LST tasks, in this paper, we propose a simple but versatile approach named Pseudo-Bidirectional Decoding (PBD) for LST tasks. PBD copies the corresponding representation of source tokens to the decoder as pseudo future context to enable the decoder to attends to its bi-directional context. In addition, the bidirectional decoding scheme and the characteristic of LST tasks motivate us to share the encoder and the decoder of seq2seq models. The proposed PBD approach provides right side context information for the decoder and models the inductive bias of LST tasks, reducing the number of parameters by half and providing good regularization effects. Experimental results on several benchmark datasets show that our approach consistently improves the performance of standard seq2seq models on LST tasks.

Wangchunshu Zhou, Tao Ge, Ke Xu et al.

Conventional Generative Adversarial Networks (GANs) for text generation tend to have issues of reward sparsity and mode collapse that affect the quality and diversity of generated samples. To address the issues, we propose a novel self-adversarial learning (SAL) paradigm for improving GANs' performance in text generation. In contrast to standard GANs that use a binary classifier as its discriminator to predict whether a sample is real or generated, SAL employs a comparative discriminator which is a pairwise classifier for comparing the text quality between a pair of samples. During training, SAL rewards the generator when its currently generated sentence is found to be better than its previously generated samples. This self-improvement reward mechanism allows the model to receive credits more easily and avoid collapsing towards the limited number of real samples, which not only helps alleviate the reward sparsity issue but also reduces the risk of mode collapse. Experiments on text generation benchmark datasets show that our proposed approach substantially improves both the quality and the diversity, and yields more stable performance compared to the previous GANs for text generation.

Yinuo Guo, Tao Ge, Furu Wei

Sentence Split and Rephrase aims to break down a complex sentence into several simple sentences with its meaning preserved. Previous studies tend to address the issue by seq2seq learning from parallel sentence pairs, which takes a complex sentence as input and sequentially generates a series of simple sentences. However, the conventional seq2seq learning has two limitations for this task: (1) it does not take into account the facts stated in the long sentence; As a result, the generated simple sentences may miss or inaccurately state the facts in the original sentence. (2) The order variance of the simple sentences to be generated may confuse the seq2seq model during training because the simple sentences derived from the long source sentence could be in any order. To overcome the challenges, we first propose the Fact-aware Sentence Encoding, which enables the model to learn facts from the long sentence and thus improves the precision of sentence split; then we introduce Permutation Invariant Training to alleviate the effects of order variance in seq2seq learning for this task. Experiments on the WebSplit-v1.0 benchmark dataset show that our approaches can largely improve the performance over the previous seq2seq learning approaches. Moreover, an extrinsic evaluation on oie-benchmark verifies the effectiveness of our approaches by an observation that splitting long sentences with our state-of-the-art model as preprocessing is helpful for improving OpenIE performance.

Wangchunshu Zhou, Tao Ge, Chang Mu et al.

We propose a novel data synthesis method to generate diverse error-corrected sentence pairs for improving grammatical error correction, which is based on a pair of machine translation models of different qualities (i.e., poor and good). The poor translation model resembles the ESL (English as a second language) learner and tends to generate translations of low quality in terms of fluency and grammatical correctness, while the good translation model generally generates fluent and grammatically correct translations. We build the poor and good translation model with phrase-based statistical machine translation model with decreased language model weight and neural machine translation model respectively. By taking the pair of their translations of the same sentences in a bridge language as error-corrected sentence pairs, we can construct unlimited pseudo parallel data. Our approach is capable of generating diverse fluency-improving patterns without being limited by the pre-defined rule set and the seed error-corrected data. Experimental results demonstrate the effectiveness of our approach and show that it can be combined with other synthetic data sources to yield further improvements.

Yi Zhang, Tao Ge, Furu Wei et al.

We study sequence-to-sequence (seq2seq) pre-training with data augmentation for sentence rewriting. Instead of training a seq2seq model with gold training data and augmented data simultaneously, we separate them to train in different phases: pre-training with the augmented data and fine-tuning with the gold data. We also introduce multiple data augmentation methods to help model pre-training for sentence rewriting. We evaluate our approach in two typical well-defined sentence rewriting tasks: Grammatical Error Correction (GEC) and Formality Style Transfer (FST). Experiments demonstrate our approach can better utilize augmented data without hurting the model's trust in gold data and further improve the model's performance with our proposed data augmentation methods. Our approach substantially advances the state-of-the-art results in well-recognized sentence rewriting benchmarks over both GEC and FST. Specifically, it pushes the CoNLL-2014 benchmark's $F_{0.5}$ score and JFLEG Test GLEU score to 62.61 and 63.54 in the restricted training setting, 66.77 and 65.22 respectively in the unrestricted setting, and advances GYAFC benchmark's BLEU to 74.24 (2.23 absolute improvement) in E&M domain and 77.97 (2.64 absolute improvement) in F&R domain.

Ruochen Xu, Tao Ge, Furu Wei

Formality style transformation is the task of modifying the formality of a given sentence without changing its content. Its challenge is the lack of large-scale sentence-aligned parallel data. In this paper, we propose an omnivorous model that takes parallel data and formality-classified data jointly to alleviate the data sparsity issue. We empirically demonstrate the effectiveness of our approach by achieving the state-of-art performance on a recently proposed benchmark dataset of formality transfer. Furthermore, our model can be readily adapted to other unsupervised text style transfer tasks like unsupervised sentiment transfer and achieve competitive results on three widely recognized benchmarks.

Tao Ge, Furu Wei, Ming Zhou

Neural sequence-to-sequence (seq2seq) approaches have proven to be successful in grammatical error correction (GEC). Based on the seq2seq framework, we propose a novel fluency boost learning and inference mechanism. Fluency boosting learning generates diverse error-corrected sentence pairs during training, enabling the error correction model to learn how to improve a sentence's fluency from more instances, while fluency boosting inference allows the model to correct a sentence incrementally with multiple inference steps. Combining fluency boost learning and inference with convolutional seq2seq models, our approach achieves the state-of-the-art performance: 75.72 (F_{0.5}) on CoNLL-2014 10 annotation dataset and 62.42 (GLEU) on JFLEG test set respectively, becoming the first GEC system that reaches human-level performance (72.58 for CoNLL and 62.37 for JFLEG) on both of the benchmarks.

Tao Ge, Qing Dou, Xiaoman Pan et al.

Aligning coordinated text streams from multiple sources and multiple languages has opened many new research venues on cross-lingual knowledge discovery. In this paper we aim to advance state-of-the-art by: (1). extending coarse-grained topic-level knowledge mining to fine-grained information units such as entities and events; (2). following a novel Data-to-Network-to-Knowledge (D2N2K) paradigm to construct and utilize network structures to capture and propagate reliable evidence. We introduce a novel Burst Information Network (BINet) representation that can display the most important information and illustrate the connections among bursty entities, events and keywords in the corpus. We propose an effective approach to construct and decipher BINets, incorporating novel criteria based on multi-dimensional clues from pronunciation, translation, burst, neighbor and graph topological structure. The experimental results on Chinese and English coordinated text streams show that our approach can accurately decipher the nodes with high confidence in the BINets and that the algorithm can be efficiently run in parallel, which makes it possible to apply it to huge amounts of streaming data for never-ending language and information decipherment.