Hwee Tou Ng

Qingyu Tan, Ruidan He, Lidong Bing et al.

Document-level Relation Extraction (DocRE) is a more challenging task compared to its sentence-level counterpart. It aims to extract relations from multiple sentences at once. In this paper, we propose a semi-supervised framework for DocRE with three novel components. Firstly, we use an axial attention module for learning the interdependency among entity-pairs, which improves the performance on two-hop relations. Secondly, we propose an adaptive focal loss to tackle the class imbalance problem of DocRE. Lastly, we use knowledge distillation to overcome the differences between human annotated data and distantly supervised data. We conducted experiments on two DocRE datasets. Our model consistently outperforms strong baselines and its performance exceeds the previous SOTA by 1.36 F1 and 1.46 Ign_F1 score on the DocRED leaderboard. Our code and data will be released at https://github.com/tonytan48/KD-DocRE.

Qingyu Tan, Lu Xu, Lidong Bing et al.

The DocRED dataset is one of the most popular and widely used benchmarks for document-level relation extraction (RE). It adopts a recommend-revise annotation scheme so as to have a large-scale annotated dataset. However, we find that the annotation of DocRED is incomplete, i.e., false negative samples are prevalent. We analyze the causes and effects of the overwhelming false negative problem in the DocRED dataset. To address the shortcoming, we re-annotate 4,053 documents in the DocRED dataset by adding the missed relation triples back to the original DocRED. We name our revised DocRED dataset Re-DocRED. We conduct extensive experiments with state-of-the-art neural models on both datasets, and the experimental results show that the models trained and evaluated on our Re-DocRED achieve performance improvements of around 13 F1 points. Moreover, we conduct a comprehensive analysis to identify the potential areas for further improvement. Our dataset is publicly available at https://github.com/tonytan48/Re-DocRED.

Qingyu Tan, Hwee Tou Ng, Lidong Bing

Reasoning about time is of fundamental importance. Many facts are time-dependent. For example, athletes change teams from time to time, and different government officials are elected periodically. Previous time-dependent question answering (QA) datasets tend to be biased in either their coverage of time spans or question types. In this paper, we introduce a comprehensive probing dataset \tempreason to evaluate the temporal reasoning capability of large language models. Our dataset includes questions of three temporal reasoning levels. In addition, we also propose a novel learning framework to improve the temporal reasoning capability of large language models, based on temporal span extraction and time-sensitive reinforcement learning. We conducted experiments in closed book QA, open book QA, and reasoning QA settings and demonstrated the effectiveness of our approach. Our code and data are released on https://github.com/DAMO-NLP-SG/TempReason.

Qingyu Tan, Hwee Tou Ng, Lidong Bing

Knowledge in the real world is being updated constantly. However, it is costly to frequently update large language models (LLMs). Therefore, it is crucial for LLMs to understand the concept of temporal knowledge. However, prior works on temporal question answering (TQA) did not emphasize multi-answer and multi-hop types of temporal reasoning. In this paper, we propose a complex temporal question-answering dataset Complex-TR that focuses on multi-answer and multi-hop temporal reasoning. Besides, we also propose a novel data augmentation strategy to improve the complex temporal reasoning capability and robustness of LLMs. We conducted experiments on multiple temporal QA datasets. Experimental results show that our method is able to improve LLMs' performance on temporal QA benchmarks by significant margins. Our code and data are released at: https://github.com/nusnlp/complex-tr.

Qingyu Tan, Lu Xu, Lidong Bing et al.

Relation extraction (RE) aims to extract relations from sentences and documents. Existing relation extraction models typically rely on supervised machine learning. However, recent studies showed that many RE datasets are incompletely annotated. This is known as the false negative problem in which valid relations are falsely annotated as 'no_relation'. Models trained with such data inevitably make similar mistakes during the inference stage. Self-training has been proven effective in alleviating the false negative problem. However, traditional self-training is vulnerable to confirmation bias and exhibits poor performance in minority classes. To overcome this limitation, we proposed a novel class-adaptive re-sampling self-training framework. Specifically, we re-sampled the pseudo-labels for each class by precision and recall scores. Our re-sampling strategy favored the pseudo-labels of classes with high precision and low recall, which improved the overall recall without significantly compromising precision. We conducted experiments on document-level and biomedical relation extraction datasets, and the results showed that our proposed self-training framework consistently outperforms existing competitive methods on the Re-DocRED and ChemDisgene datasets when the training data are incompletely annotated. Our code is released at https://github.com/DAMO-NLP-SG/CAST.

Christopher Bryant, Zheng Yuan, Muhammad Reza Qorib et al.

Grammatical Error Correction (GEC) is the task of automatically detecting and correcting errors in text. The task not only includes the correction of grammatical errors, such as missing prepositions and mismatched subject-verb agreement, but also orthographic and semantic errors, such as misspellings and word choice errors respectively. The field has seen significant progress in the last decade, motivated in part by a series of five shared tasks, which drove the development of rule-based methods, statistical classifiers, statistical machine translation, and finally neural machine translation systems which represent the current dominant state of the art. In this survey paper, we condense the field into a single article and first outline some of the linguistic challenges of the task, introduce the most popular datasets that are available to researchers (for both English and other languages), and summarise the various methods and techniques that have been developed with a particular focus on artificial error generation. We next describe the many different approaches to evaluation as well as concerns surrounding metric reliability, especially in relation to subjective human judgements, before concluding with an overview of recent progress and suggestions for future work and remaining challenges. We hope that this survey will serve as comprehensive resource for researchers who are new to the field or who want to be kept apprised of recent developments.

Hai Ye, Hwee Tou Ng

Pre-trained large language models (LLMs) can be tailored to adhere to human instructions through instruction tuning. However, due to shifts in the distribution of test-time data, they may not always execute instructions accurately, potentially generating factual errors or misaligned content when acting as chat assistants. To enhance the reliability of LLMs in following instructions, we propose the study of selective instruction following, whereby the system declines to execute instructions if the anticipated response quality is low. We train judge models that can predict numerical quality scores for model responses. To address data scarcity, we introduce Self-J, a novel self-training framework for developing judge models without needing human-annotated quality scores. Our method leverages the model's inherent self-evaluation capability to extract information about response quality from labeled instruction-tuning data. It incorporates a gold reference answer to facilitate self-evaluation and recalibrates by assessing the semantic similarity between the response sample and the gold reference. During the training phase, we implement self-distillation as a regularization technique to enhance the capability of reference-free estimation. To validate alignment evaluation on general instruction-following tasks, we collect large-scale high-quality instructions from Hugging Face for model training and evaluation. Extensive experiments on five open-source models show that our method correlates much more with GPT-4 than strong baselines, e.g., supervised models distilled from GPT-4 and GPT-3.5-turbo. Our analysis shows our model's strong generalization across domains. Additionally, our judge models serve as good reward models, e.g., boosting WizardLM-13B-V1.2 from 89.17 to 92.48 and from 12.03 to 15.90 in version v1 and v2 of AlpacaEval respectively using best-of-32 sampling with our judge models.

Hai Ye, Yuyang Ding, Juntao Li et al.

A deployed question answering (QA) model can easily fail when the test data has a distribution shift compared to the training data. Robustness tuning (RT) methods have been widely studied to enhance model robustness against distribution shifts before model deployment. However, can we improve a model after deployment? To answer this question, we evaluate test-time adaptation (TTA) to improve a model after deployment. We first introduce COLDQA, a unified evaluation benchmark for robust QA against text corruption and changes in language and domain. We then evaluate previous TTA methods on COLDQA and compare them to RT methods. We also propose a novel TTA method called online imitation learning (OIL). Through extensive experiments, we find that TTA is comparable to RT methods, and applying TTA after RT can significantly boost the performance on COLDQA. Our proposed OIL improves TTA to be more robust to variation in hyper-parameters and test distributions over time.

Hai Ye, Qizhe Xie, Hwee Tou Ng

In this work, we study multi-source test-time model adaptation from user feedback, where K distinct models are established for adaptation. To allow efficient adaptation, we cast the problem as a stochastic decision-making process, aiming to determine the best adapted model after adaptation. We discuss two frameworks: multi-armed bandit learning and multi-armed dueling bandits. Compared to multi-armed bandit learning, the dueling framework allows pairwise collaboration among K models, which is solved by a novel method named Co-UCB proposed in this work. Experiments on six datasets of extractive question answering (QA) show that the dueling framework using Co-UCB is more effective than other strong baselines for our studied problem.

Hai Ye, Hwee Tou Ng, Wenjuan Han

In conversational question answering (CQA), the task of question rewriting~(QR) in context aims to rewrite a context-dependent question into an equivalent self-contained question that gives the same answer. In this paper, we are interested in the robustness of a QR system to questions varying in rewriting hardness or difficulty. Since there is a lack of questions classified based on their rewriting hardness, we first propose a heuristic method to automatically classify questions into subsets of varying hardness, by measuring the discrepancy between a question and its rewrite. To find out what makes questions hard or easy for rewriting, we then conduct a human evaluation to annotate the rewriting hardness of questions. Finally, to enhance the robustness of QR systems to questions of varying hardness, we propose a novel learning framework for QR that first trains a QR model independently on each subset of questions of a certain level of hardness, then combines these QR models as one joint model for inference. Experimental results on two datasets show that our framework improves the overall performance compared to the baselines.

Wenzheng Zeng, Difei Gao, Mike Zheng Shou et al.

Recent video-language models have shown great potential for video understanding, but still struggle with accurate temporal grounding for event-level perception. We observe that two main factors in video understanding (i.e., temporal grounding and textual response) form a logical hierarchy: accurate temporal evidence grounding lays the foundation for reliable textual response. However, existing works typically handle these two tasks in a coupled manner without a clear logical structure, leading to sub-optimal objectives. We address this from a factorized learning perspective. We first propose D$^2$VLM, a framework that decouples the learning of these two tasks while also emphasizing their inherent dependency. We adopt a "grounding then answering with evidence referencing" paradigm and introduce evidence tokens for evidence grounding, which emphasize event-level visual semantic capture beyond the focus on timestamp representation in existing works. To further facilitate the learning of these two tasks, we introduce a novel factorized preference optimization (FPO) algorithm. Unlike standard preference optimization, FPO explicitly incorporates probabilistic temporal grounding modeling into the optimization objective, enabling preference learning for both temporal grounding and textual response. We also construct a synthetic dataset to address the lack of suitable datasets for factorized preference learning with explicit temporal grounding. Experiments on various tasks demonstrate the clear advantage of our approach. Our source code is available at https://github.com/nusnlp/d2vlm.

Tan Sang Nguyen, Muhammad Reza Qorib, Hwee Tou Ng

Large language models (LLMs) have proven to be effective tools for a wide range of natural language processing (NLP) applications. Although many LLMs are multilingual, most remain English-centric and perform poorly on low-resource languages. Recently, several Southeast Asia-focused LLMs have been developed, but none are truly open source, as they do not publicly disclose their training data. Truly open-source models are important for transparency and for enabling a deeper and more precise understanding of LLM internals and development, including biases, generalization, and multilinguality. Motivated by recent advances demonstrating the effectiveness of parallel data in improving multilingual performance, we conduct controlled and comprehensive experiments to study the effectiveness of parallel data in continual pretraining of LLMs. Our findings show that using only parallel data is the most effective way to extend an LLM to new languages. Using just 34.7B tokens of parallel data and 180 hours on 8x NVIDIA H200 GPUs, we built OpenSeal, the first truly open Southeast Asian LLM that rivals the performance of existing models of similar size.

Hai Ye, Hwee Tou Ng

Iterative data generation and model re-training can effectively align large language models(LLMs) to human preferences. The process of data sampling is crucial, as it significantly influences the success of policy improvement. Repeated random sampling is a widely used method that independently queries the model multiple times to generate outputs. In this work, we propose a more effective sampling method, named Preference-Guided Reflective Sampling (PRS). Unlike random sampling, PRS employs a tree-based generation framework to enable more efficient sampling. It leverages adaptive self-refinement techniques to better explore the sampling space. By specifying user preferences in natural language, PRS can further optimize response generation according to these preferences. As a result, PRS can align models to diverse user preferences. Our experiments demonstrate that PRS generates higher-quality responses with significantly higher rewards. On AlpacaEval and Arena-Hard, PRS substantially outperforms repeated random sampling in best-of-$N$ sampling. Moreover, PRS shows strong performance when applied in iterative offline RL training.

Mingyu Ouyang, Siyuan Hu, Kevin Qinghong Lin et al.

Towards an embodied generalist for real-world interaction, Multimodal Large Language Model (MLLM) agents still suffer from challenging latency, sparse feedback, and irreversible mistakes. Video games offer an ideal testbed with rich visual observations and closed-loop interaction, demanding fine-grained perception, long-horizon planning, and precise control. However, systematically evaluating these capabilities is currently hindered by heterogeneous action interfaces and heuristic verification. To this end, we introduce GameWorld, a benchmark designed for standardized and verifiable evaluation of MLLMs as generalist game agents in browser environments. Two game agent interfaces are studied: (i) computer-use agents that directly emit keyboard and mouse controls, and (ii) generalist multimodal agents that act in a semantic action space via deterministic Semantic Action Parsing. GameWorld contains 34 diverse games and 170 tasks, each paired with state-verifiable metrics for outcome-based evaluation. The results across 18 model-interface pairs suggest that even the best performing agent is far from achieving human capabilities on video games. Extensive experiments of repeated full-benchmark reruns demonstrate the robustness of the benchmark, while further studies on real-time interaction, context-memory sensitivity, and action validity expose more challenges ahead for game agents. Together, by offering a standardized, verifiable, and reproducible evaluation framework, GameWorld lays a robust foundation for advancing research on multimodal game agents and beyond. The project page is at https://gameworld-bench.github.io.

Muhammad Reza Qorib, Hwee Tou Ng

Quality estimation models have been developed to assess the corrections made by grammatical error correction (GEC) models when the reference or gold-standard corrections are not available. An ideal quality estimator can be utilized to combine the outputs of multiple GEC systems by choosing the best subset of edits from the union of all edits proposed by the GEC base systems. However, we found that existing GEC quality estimation models are not good enough in differentiating good corrections from bad ones, resulting in a low F0.5 score when used for system combination. In this paper, we propose GRECO, a new state-of-the-art quality estimation model that gives a better estimate of the quality of a corrected sentence, as indicated by having a higher correlation to the F0.5 score of a corrected sentence. It results in a combined GEC system with a higher F0.5 score. We also propose three methods for utilizing GEC quality estimation models for system combination with varying generality: model-agnostic, model-agnostic with voting bias, and model-dependent method. The combined GEC system outperforms the state of the art on the CoNLL-2014 test set and the BEA-2019 test set, achieving the highest F0.5 scores published to date.

Mingyu Ouyang, Kevin Qinghong Lin, Mike Zheng Shou et al.

Vision-Language Models (VLMs) have shown remarkable performance in User Interface (UI) grounding tasks, driven by their ability to process increasingly high-resolution screenshots. However, screenshots are tokenized into thousands of visual tokens (e.g., about 4700 for 2K resolution), incurring significant computational overhead and diluting attention. In contrast, humans typically focus on regions of interest when interacting with UI. In this work, we pioneer the task of efficient UI grounding. Guided by practical analysis of the task's characteristics and challenges, we propose FocusUI, an efficient UI grounding framework that selects patches most relevant to the instruction while preserving positional continuity for precise grounding. FocusUI addresses two key challenges: (1) Eliminating redundant tokens in visual encoding. We construct patch-level supervision by fusing an instruction-conditioned score with a rule-based UI-graph score that down-weights large homogeneous regions to select distinct and instruction-relevant visual tokens. (2) Preserving positional continuity during visual token selection. We find that general visual token pruning methods suffer from severe accuracy degradation on UI grounding tasks due to broken positional information. We introduce a novel PosPad strategy, which compresses each contiguous sequence of dropped visual tokens into a single special marker placed at the sequence's last index to preserve positional continuity. Comprehensive experiments on four grounding benchmarks demonstrate that FocusUI surpasses GUI-specific baselines. On the ScreenSpot-Pro benchmark, FocusUI-7B achieves a performance improvement of 3.7% over GUI-Actor-7B. Even with only 30% visual token retention, FocusUI-7B drops by only 3.2% while achieving up to 1.44x faster inference and 17% lower peak GPU memory.

Christopher Adrian Kusuma, Muhammad Reza Qorib, Hwee Tou Ng

Large language models (LLMs) are highly capable of answering questions, but they are often unaware of their own knowledge boundary, i.e., knowing what they know and what they don't know. As a result, they can generate factually incorrect responses on topics they do not have enough knowledge of, commonly known as hallucination. Rather than hallucinating, a language model should be more honest and respond with "I don't know" when it does not have enough knowledge about a topic. Many methods have been proposed to improve LLM honesty, but their evaluations lack robustness, as they do not take into account the knowledge that the LLM has ingested during its pretraining. In this paper, we propose a more robust evaluation benchmark dataset for LLM honesty by utilizing Pythia, a truly open LLM with publicly available pretraining data. In addition, we also propose a novel method for harnessing the pretraining data to build a more honest LLM.

Wenjuan Han, Hwee Tou Ng

Grammatical error correction (GEC) is the task of detecting and correcting errors in a written text. The idea of combining multiple system outputs has been successfully used in GEC. To achieve successful system combination, multiple component systems need to produce corrected sentences that are both diverse and of comparable quality. However, most existing state-of-the-art GEC approaches are based on similar sequence-to-sequence neural networks, so the gains are limited from combining the outputs of component systems similar to one another. In this paper, we present Diversity-Driven Combination (DDC) for GEC, a system combination strategy that encourages diversity among component systems. We evaluate our system combination strategy on the CoNLL-2014 shared task and the BEA-2019 shared task. On both benchmarks, DDC achieves significant performance gain with a small number of training examples and outperforms the component systems by a large margin. Our source code is available at https://github.com/nusnlp/gec-ddc.

Juntao Li, Ruidan He, Hai Ye et al.

Recent research indicates that pretraining cross-lingual language models on large-scale unlabeled texts yields significant performance improvements over various cross-lingual and low-resource tasks. Through training on one hundred languages and terabytes of texts, cross-lingual language models have proven to be effective in leveraging high-resource languages to enhance low-resource language processing and outperform monolingual models. In this paper, we further investigate the cross-lingual and cross-domain (CLCD) setting when a pretrained cross-lingual language model needs to adapt to new domains. Specifically, we propose a novel unsupervised feature decomposition method that can automatically extract domain-specific features and domain-invariant features from the entangled pretrained cross-lingual representations, given unlabeled raw texts in the source language. Our proposed model leverages mutual information estimation to decompose the representations computed by a cross-lingual model into domain-invariant and domain-specific parts. Experimental results show that our proposed method achieves significant performance improvements over the state-of-the-art pretrained cross-lingual language model in the CLCD setting. The source code of this paper is publicly available at https://github.com/lijuntaopku/UFD.

Christian Hadiwinoto, Hwee Tou Ng, Wee Chung Gan

Contextualized word representations are able to give different representations for the same word in different contexts, and they have been shown to be effective in downstream natural language processing tasks, such as question answering, named entity recognition, and sentiment analysis. However, evaluation on word sense disambiguation (WSD) in prior work shows that using contextualized word representations does not outperform the state-of-the-art approach that makes use of non-contextualized word embeddings. In this paper, we explore different strategies of integrating pre-trained contextualized word representations and our best strategy achieves accuracies exceeding the best prior published accuracies by significant margins on multiple benchmark WSD datasets. We make the source code available at https://github.com/nusnlp/contextemb-wsd.

Wenzheng Zeng, Mingyu Ouyang, Langyuan Cui et al.

Automatic presentation slide generation can greatly streamline content creation. However, since preferences of each user may vary, existing under-specified formulations often lead to suboptimal results that fail to align with individual user needs. We introduce a novel task that conditions paper-to-slides generation on user-specified preferences. We propose a human behavior-inspired agentic framework, SlideTailor, that progressively generates editable slides in a user-aligned manner. Instead of requiring users to write their preferences in detailed textual form, our system only asks for a paper-slides example pair and a visual template - natural and easy-to-provide artifacts that implicitly encode rich user preferences across content and visual style. Despite the implicit and unlabeled nature of these inputs, our framework effectively distills and generalizes the preferences to guide customized slide generation. We also introduce a novel chain-of-speech mechanism to align slide content with planned oral narration. Such a design significantly enhances the quality of generated slides and enables downstream applications like video presentations. To support this new task, we construct a benchmark dataset that captures diverse user preferences, with carefully designed interpretable metrics for robust evaluation. Extensive experiments demonstrate the effectiveness of our framework.

Hai Ye, Mingbao Lin, Hwee Tou Ng et al.

Scaling laws for inference compute in multi-agent systems remain under-explored compared to single-agent scenarios. This work aims to bridge this gap by investigating the problem of data synthesis through multi-agent sampling, where synthetic responses are generated by sampling from multiple distinct language models. Effective model coordination is crucial for successful multi-agent collaboration. Unlike previous approaches that rely on fixed workflows, we treat model coordination as a multi-step decision-making process, optimizing generation structures dynamically for each input question. We introduce Tree Search-based Orchestrated Agents~(TOA), where the workflow evolves iteratively during the sequential sampling process. To achieve this, we leverage Monte Carlo Tree Search (MCTS), integrating a reward model to provide real-time feedback and accelerate exploration. Our experiments on alignment, machine translation, and mathematical reasoning demonstrate that multi-agent sampling significantly outperforms single-agent sampling as inference compute scales. TOA is the most compute-efficient approach, achieving SOTA performance on WMT and a 72.2\% LC win rate on AlpacaEval. Moreover, fine-tuning with our synthesized alignment data surpasses strong preference learning methods on challenging benchmarks such as Arena-Hard and AlpacaEval.

Junyi Li, Hwee Tou Ng

Large language models (LLMs) have significantly advanced in reasoning tasks through reinforcement learning (RL) optimization, achieving impressive capabilities across various challenging benchmarks. However, our empirical analysis reveals a critical drawback: reasoning-oriented RL fine-tuning significantly increases the prevalence of hallucinations. We theoretically analyze the RL training dynamics, identifying high-variance gradient, entropy-induced randomness, and susceptibility to spurious local optima as key factors leading to hallucinations. To address this drawback, we propose Factuality-aware Step-wise Policy Optimization (FSPO), an innovative RL fine-tuning algorithm incorporating explicit factuality verification at each reasoning step. FSPO leverages automated verification against given evidence to dynamically adjust token-level advantage values, incentivizing factual correctness throughout the reasoning process. Experiments across mathematical reasoning and hallucination benchmarks using Qwen2.5 and Llama models demonstrate that FSPO effectively reduces hallucinations while enhancing reasoning accuracy, substantially improving both reliability and performance.

Yao Xiao, Hai Ye, Linyao Chen et al.

Iterative data generation and model retraining are widely used to align large language models (LLMs). It typically involves a policy model to generate on-policy responses and a reward model to guide training data selection. Direct Preference Optimization (DPO) further enhances this process by constructing preference pairs of chosen and rejected responses. In this work, we aim to \emph{scale up} the number of on-policy samples via repeated random sampling to improve alignment performance. Conventional practice selects the sample with the highest reward as chosen and the lowest as rejected for DPO. However, our experiments reveal that this strategy leads to a \emph{decline} in performance as the sample size increases. To address this, we investigate preference data construction through the lens of underlying normal distribution of sample rewards. We categorize the reward space into seven representative points and systematically explore all 21 ($C_7^2$) pairwise combinations. Through evaluations on four models using AlpacaEval 2, we find that selecting the rejected response at reward position $μ- 2σ$ rather than the minimum reward, is crucial for optimal performance. We finally introduce a scalable preference data construction strategy that consistently enhances model performance as the sample scale increases.

Muhammad Reza Qorib, Qisheng Hu, Hwee Tou Ng

Given news articles about an entity, such as a public figure or organization, timeline summarization (TLS) involves generating a timeline that summarizes the key events about the entity. However, the TLS task is too underspecified, since what is of interest to each reader may vary, and hence there is not a single ideal or optimal timeline. In this paper, we introduce a novel task, called Constrained Timeline Summarization (CTLS), where a timeline is generated in which all events in the timeline meet some constraint. An example of a constrained timeline concerns the legal battles of Tiger Woods, where only events related to his legal problems are selected to appear in the timeline. We collected a new human-verified dataset of constrained timelines involving 47 entities and 5 constraints per entity. We propose an approach that employs a large language model (LLM) to summarize news articles according to a specified constraint and cluster them to identify key events to include in a constrained timeline. In addition, we propose a novel self-reflection method during summary generation, demonstrating that this approach successfully leads to improved performance.

Junyi Li, Hwee Tou Ng

Despite their outstanding capabilities, large language models (LLMs) are prone to hallucination and producing factually incorrect information. This challenge has spurred efforts in attributed text generation, which prompts LLMs to generate content with supporting evidence. In this paper, we propose a novel framework, called Think&Cite, and formulate attributed text generation as a multi-step reasoning problem integrated with search. Specifically, we propose Self-Guided Monte Carlo Tree Search (SG-MCTS), which capitalizes on the self-reflection capability of LLMs to reason about the intermediate states of MCTS for guiding the tree expansion process. To provide reliable and comprehensive feedback, we introduce Progress Reward Modeling to measure the progress of tree search from the root to the current state from two aspects, i.e., generation and attribution progress. We conduct extensive experiments on three datasets and the results show that our approach significantly outperforms baseline approaches.

Muhammad Reza Qorib, Junyi Li, Hwee Tou Ng

Large language models (LLMs) have demonstrated impressive translation capabilities even without being explicitly trained on parallel data. This remarkable property has led some to believe that parallel data is no longer necessary for building multilingual language models. While some attribute this to the emergent abilities of LLMs due to scale, recent work suggests that it is actually caused by incidental bilingual signals present in the training data. Various methods have been proposed to maximize the utility of parallel data to enhance the multilingual capabilities of multilingual encoder-based and encoder-decoder language models. However, some decoder-based LLMs opt to ignore parallel data instead. In this work, we conduct a systematic study on the impact of adding parallel data on LLMs' multilingual capabilities, focusing specifically on translation and multilingual common-sense reasoning. Through controlled experiments, we demonstrate that parallel data can significantly improve LLMs' multilingual capabilities.

Muhammad Reza Qorib, Alham Fikri Aji, Hwee Tou Ng

Error type information has been widely used to improve the performance of grammatical error correction (GEC) models, whether for generating corrections, re-ranking them, or combining GEC models. Combining GEC models that have complementary strengths in correcting different error types is very effective in producing better corrections. However, system combination incurs a high computational cost due to the need to run inference on the base systems before running the combination method itself. Therefore, it would be more efficient to have a single model with multiple sub-networks that specialize in correcting different error types. In this paper, we propose a mixture-of-experts model, MoECE, for grammatical error correction. Our model successfully achieves the performance of T5-XL with three times fewer effective parameters. Additionally, our model produces interpretable corrections by also identifying the error type during inference.

Langyuan Cui, Chun Kai Ling, Hwee Tou Ng

Large Language Models (LLMs) are increasingly deployed in real-world scenarios where they may lack sufficient information to complete a given task. In such settings, the ability to actively seek out missing information becomes a critical capability. Existing approaches to enhancing this ability often rely on simplifying assumptions that degrade \textit{worst-case} performance. This is an issue with serious implications in high-stakes applications. In this work, we use the game of Twenty Questions to evaluate the information-seeking ability of LLMs. We introduce and formalize its adversarial counterpart, the Strategic Language Search (SLS) problem along with its variants as a two-player zero-sum extensive form game. We propose Game of Thought (GoT), a framework that applies game-theoretic techniques to approximate a Nash equilibrium (NE) strategy for the restricted variant of the game. Empirical results demonstrate that our approach consistently improves worst-case performance compared to (1) direct prompting-based methods and (2) heuristic-guided search methods across all tested settings.

Hwee Tou Ng, Siew Mei Wu, Yuanbin Wu et al.

The CoNLL-2013 shared task was devoted to grammatical error correction. In this paper, we give the task definition, present the data sets, and describe the evaluation metric and scorer used in the shared task. We also give an overview of the various approaches adopted by the participating teams, and present the evaluation results.

Xingxuan Li, Liying Cheng, Qingyu Tan et al.

The temporal aspect is a significant dimension of our reality. We notice the challenge that large language models (LLMs) face when engaging in temporal reasoning. Our preliminary experiments show that methods involving the generation of intermediate reasoning steps, such as chain-of-thought and program-aided language models, do not consistently boost the performance of complex temporal question-answering tasks. This limitation can be attributed to the LLMs' inadequate understanding of temporal information. To address this problem, we propose TempLogic, a novel framework designed specifically for temporal question-answering tasks across three levels of reasoning. TempLogic incorporates retrieval-guided context distillation, temporal data extraction, and tailor-made logic reasoning. Extensive experiments and analysis demonstrate the effectiveness of our framework in solving intricate time-bound reasoning tasks.

Qian Lin, Hwee Tou Ng

Identifying breakdowns in ongoing dialogues helps to improve communication effectiveness. Most prior work on this topic relies on human annotated data and data augmentation to learn a classification model. While quality labeled dialogue data requires human annotation and is usually expensive to obtain, unlabeled data is easier to collect from various sources. In this paper, we propose a novel semi-supervised teacher-student learning framework to tackle this task. We introduce two teachers which are trained on labeled data and perturbed labeled data respectively. We leverage unlabeled data to improve classification in student training where we employ two teachers to refine the labeling of unlabeled data through teacher-student learning in a bootstrapping manner. Through our proposed training approach, the student can achieve improvements over single-teacher performance. Experimental results on the Dialogue Breakdown Detection Challenge dataset DBDC5 and Learning to Identify Follow-Up Questions dataset LIF show that our approach outperforms all previous published approaches as well as other supervised and semi-supervised baseline methods.

Ruixi Lin, Hwee Tou Ng

In this paper, we propose a system combination method for grammatical error correction (GEC), based on nonlinear integer programming (IP). Our method optimizes a novel F score objective based on error types, and combines multiple end-to-end GEC systems. The proposed IP approach optimizes the selection of a single best system for each grammatical error type present in the data. Experiments of the IP approach on combining state-of-the-art standalone GEC systems show that the combined system outperforms all standalone systems. It improves F0.5 score by 3.61% when combining the two best participating systems in the BEA 2019 shared task, and achieves F0.5 score of 73.08%. We also perform experiments to compare our IP approach with another state-of-the-art system combination method for GEC, demonstrating IP's competitive combination capability.

Preslav Nakov, Hwee Tou Ng

We propose a novel approach to translating from a morphologically complex language. Unlike previous research, which has targeted word inflections and concatenations, we focus on the pairwise relationship between morphologically related words, which we treat as potential paraphrases and handle using paraphrasing techniques at the word, phrase, and sentence level. An important advantage of this framework is that it can cope with derivational morphology, which has so far remained largely beyond the capabilities of statistical machine translation systems. Our experiments translating from Malay, whose morphology is mostly derivational, into English show significant improvements over rivaling approaches based on five automatic evaluation measures (for 320,000 sentence pairs; 9.5 million English word tokens).

Tapas Nayak, Hwee Tou Ng

Distantly supervised datasets for relation extraction mostly focus on sentence-level extraction, and they cover very few relations. In this work, we propose cross-document relation extraction, where the two entities of a relation tuple appear in two different documents that are connected via a chain of common entities. Following this idea, we create a dataset for two-hop relation extraction, where each chain contains exactly two documents. Our proposed dataset covers a higher number of relations than the publicly available sentence-level datasets. We also propose a hierarchical entity graph convolutional network (HEGCN) model for this task that improves performance by 1.1\% F1 score on our two-hop relation extraction dataset, compared to some strong neural baselines.

Wenjuan Han, Yong Jiang, Hwee Tou Ng et al.

Syntactic dependency parsing is an important task in natural language processing. Unsupervised dependency parsing aims to learn a dependency parser from sentences that have no annotation of their correct parse trees. Despite its difficulty, unsupervised parsing is an interesting research direction because of its capability of utilizing almost unlimited unannotated text data. It also serves as the basis for other research in low-resource parsing. In this paper, we survey existing approaches to unsupervised dependency parsing, identify two major classes of approaches, and discuss recent trends. We hope that our survey can provide insights for researchers and facilitate future research on this topic.

Hai Ye, Qingyu Tan, Ruidan He et al.

Adapting pre-trained language models (PrLMs) (e.g., BERT) to new domains has gained much attention recently. Instead of fine-tuning PrLMs as done in most previous work, we investigate how to adapt the features of PrLMs to new domains without fine-tuning. We explore unsupervised domain adaptation (UDA) in this paper. With the features from PrLMs, we adapt the models trained with labeled data from the source domain to the unlabeled target domain. Self-training is widely used for UDA which predicts pseudo labels on the target domain data for training. However, the predicted pseudo labels inevitably include noise, which will negatively affect training a robust model. To improve the robustness of self-training, in this paper we present class-aware feature self-distillation (CFd) to learn discriminative features from PrLMs, in which PrLM features are self-distilled into a feature adaptation module and the features from the same class are more tightly clustered. We further extend CFd to a cross-language setting, in which language discrepancy is studied. Experiments on two monolingual and multilingual Amazon review datasets show that CFd can consistently improve the performance of self-training in cross-domain and cross-language settings.

Yixuan Tang, Hwee Tou Ng, Anthony K. H. Tung

Multi-hop question answering (QA) requires a model to retrieve and integrate information from different parts of a long text to answer a question. Humans answer this kind of complex questions via a divide-and-conquer approach. In this paper, we investigate whether top-performing models for multi-hop questions understand the underlying sub-questions like humans. We adopt a neural decomposition model to generate sub-questions for a multi-hop complex question, followed by extracting the corresponding sub-answers. We show that multiple state-of-the-art multi-hop QA models fail to correctly answer a large portion of sub-questions, although their corresponding multi-hop questions are correctly answered. This indicates that these models manage to answer the multi-hop questions using some partial clues, instead of truly understanding the reasoning paths. We also propose a new model which significantly improves the performance on answering the sub-questions. Our work takes a step forward towards building a more explainable multi-hop QA system.

Tapas Nayak, Hwee Tou Ng

Relation extraction is the task of determining the relation between two entities in a sentence. Distantly-supervised models are popular for this task. However, sentences can be long and two entities can be located far from each other in a sentence. The pieces of evidence supporting the presence of a relation between two entities may not be very direct, since the entities may be connected via some indirect links such as a third entity or via co-reference. Relation extraction in such scenarios becomes more challenging as we need to capture the long-distance interactions among the entities and other words in the sentence. Also, the words in a sentence do not contribute equally in identifying the relation between the two entities. To address this issue, we propose a novel and effective attention model which incorporates syntactic information of the sentence and a multi-factor attention mechanism. Experiments on the New York Times corpus show that our proposed model outperforms prior state-of-the-art models.

Tapas Nayak, Hwee Tou Ng

A relation tuple consists of two entities and the relation between them, and often such tuples are found in unstructured text. There may be multiple relation tuples present in a text and they may share one or both entities among them. Extracting such relation tuples from a sentence is a difficult task and sharing of entities or overlapping entities among the tuples makes it more challenging. Most prior work adopted a pipeline approach where entities were identified first followed by finding the relations among them, thus missing the interaction among the relation tuples in a sentence. In this paper, we propose two approaches to use encoder-decoder architecture for jointly extracting entities and relations. In the first approach, we propose a representation scheme for relation tuples which enables the decoder to generate one word at a time like machine translation models and still finds all the tuples present in a sentence with full entity names of different length and with overlapping entities. Next, we propose a pointer network-based decoding approach where an entire tuple is generated at every time step. Experiments on the publicly available New York Times corpus show that our proposed approaches outperform previous work and achieve significantly higher F1 scores.

Ruidan He, Wee Sun Lee, Hwee Tou Ng et al.

Aspect-based sentiment analysis produces a list of aspect terms and their corresponding sentiments for a natural language sentence. This task is usually done in a pipeline manner, with aspect term extraction performed first, followed by sentiment predictions toward the extracted aspect terms. While easier to develop, such an approach does not fully exploit joint information from the two subtasks and does not use all available sources of training information that might be helpful, such as document-level labeled sentiment corpus. In this paper, we propose an interactive multi-task learning network (IMN) which is able to jointly learn multiple related tasks simultaneously at both the token level as well as the document level. Unlike conventional multi-task learning methods that rely on learning common features for the different tasks, IMN introduces a message passing architecture where information is iteratively passed to different tasks through a shared set of latent variables. Experimental results demonstrate superior performance of the proposed method against multiple baselines on three benchmark datasets.

Ruidan He, Wee Sun Lee, Hwee Tou Ng et al.

We consider the cross-domain sentiment classification problem, where a sentiment classifier is to be learned from a source domain and to be generalized to a target domain. Our approach explicitly minimizes the distance between the source and the target instances in an embedded feature space. With the difference between source and target minimized, we then exploit additional information from the target domain by consolidating the idea of semi-supervised learning, for which, we jointly employ two regularizations -- entropy minimization and self-ensemble bootstrapping -- to incorporate the unlabeled target data for classifier refinement. Our experimental results demonstrate that the proposed approach can better leverage unlabeled data from the target domain and achieve substantial improvements over baseline methods in various experimental settings.

Ruidan He, Wee Sun Lee, Hwee Tou Ng et al.

Attention-based long short-term memory (LSTM) networks have proven to be useful in aspect-level sentiment classification. However, due to the difficulties in annotating aspect-level data, existing public datasets for this task are all relatively small, which largely limits the effectiveness of those neural models. In this paper, we explore two approaches that transfer knowledge from document- level data, which is much less expensive to obtain, to improve the performance of aspect-level sentiment classification. We demonstrate the effectiveness of our approaches on 4 public datasets from SemEval 2014, 2015, and 2016, and we show that attention-based LSTM benefits from document-level knowledge in multiple ways.

Christian Hadiwinoto, Hwee Tou Ng

There are many machine translation (MT) papers that propose novel approaches and show improvements over their self-defined baselines. The experimental setting in each paper often differs from one another. As such, it is hard to determine if a proposed approach is really useful and advances the state of the art. Chinese-to-English translation is a common translation direction in MT papers, although there is not one widely accepted experimental setting in Chinese-to-English MT. Our goal in this paper is to propose a benchmark in evaluation setup for Chinese-to-English machine translation, such that the effectiveness of a new proposed MT approach can be directly compared to previous approaches. Towards this end, we also built a highly competitive state-of-the-art MT system trained on a large-scale training set. Our system outperforms reported results on NIST OpenMT test sets in almost all papers published in major conferences and journals in computational linguistics and artificial intelligence in the past 11 years. We argue that a standardized benchmark on data and performance is important for meaningful comparison.

Shamil Chollampatt, Hwee Tou Ng

We improve automatic correction of grammatical, orthographic, and collocation errors in text using a multilayer convolutional encoder-decoder neural network. The network is initialized with embeddings that make use of character N-gram information to better suit this task. When evaluated on common benchmark test data sets (CoNLL-2014 and JFLEG), our model substantially outperforms all prior neural approaches on this task as well as strong statistical machine translation-based systems with neural and task-specific features trained on the same data. Our analysis shows the superiority of convolutional neural networks over recurrent neural networks such as long short-term memory (LSTM) networks in capturing the local context via attention, and thereby improving the coverage in correcting grammatical errors. By ensembling multiple models, and incorporating an N-gram language model and edit features via rescoring, our novel method becomes the first neural approach to outperform the current state-of-the-art statistical machine translation-based approach, both in terms of grammaticality and fluency.

Souvik Kundu, Hwee Tou Ng

Neural network models recently proposed for question answering (QA) primarily focus on capturing the passage-question relation. However, they have minimal capability to link relevant facts distributed across multiple sentences which is crucial in achieving deeper understanding, such as performing multi-sentence reasoning, co-reference resolution, etc. They also do not explicitly focus on the question and answer type which often plays a critical role in QA. In this paper, we propose a novel end-to-end question-focused multi-factor attention network for answer extraction. Multi-factor attentive encoding using tensor-based transformation aggregates meaningful facts even when they are located in multiple sentences. To implicitly infer the answer type, we also propose a max-attentional question aggregation mechanism to encode a question vector based on the important words in a question. During prediction, we incorporate sequence-level encoding of the first wh-word and its immediately following word as an additional source of question type information. Our proposed model achieves significant improvements over the best prior state-of-the-art results on three large-scale challenging QA datasets, namely NewsQA, TriviaQA, and SearchQA.

Christian Hadiwinoto, Hwee Tou Ng

In machine translation (MT) that involves translating between two languages with significant differences in word order, determining the correct word order of translated words is a major challenge. The dependency parse tree of a source sentence can help to determine the correct word order of the translated words. In this paper, we present a novel reordering approach utilizing a neural network and dependency-based embeddings to predict whether the translations of two source words linked by a dependency relation should remain in the same order or should be swapped in the translated sentence. Experiments on Chinese-to-English translation show that our approach yields a statistically significant improvement of 0.57 BLEU point on benchmark NIST test sets, compared to our prior state-of-the-art statistical MT system that uses sparse dependency-based reordering features.

Christian Hadiwinoto, Yang Liu, Hwee Tou Ng

Reordering poses a major challenge in machine translation (MT) between two languages with significant differences in word order. In this paper, we present a novel reordering approach utilizing sparse features based on dependency word pairs. Each instance of these features captures whether two words, which are related by a dependency link in the source sentence dependency parse tree, follow the same order or are swapped in the translation output. Experiments on Chinese-to-English translation show a statistically significant improvement of 1.21 BLEU point using our approach, compared to a state-of-the-art statistical MT system that incorporates prior reordering approaches.

Duc Tam Hoang, Shamil Chollampatt, Hwee Tou Ng

Grammatical error correction (GEC) is the task of detecting and correcting grammatical errors in texts written by second language learners. The statistical machine translation (SMT) approach to GEC, in which sentences written by second language learners are translated to grammatically correct sentences, has achieved state-of-the-art accuracy. However, the SMT approach is unable to utilize global context. In this paper, we propose a novel approach to improve the accuracy of GEC, by exploiting the n-best hypotheses generated by an SMT approach. Specifically, we build a classifier to score the edits in the n-best hypotheses. The classifier can be used to select appropriate edits or re-rank the n-best hypotheses. We apply these methods to a state-of-the-art GEC system that uses the SMT approach. Our experiments show that our methods achieve statistically significant improvements in accuracy over the best published results on a benchmark test dataset on GEC.

Shamil Chollampatt, Kaveh Taghipour, Hwee Tou Ng

Phrase-based statistical machine translation (SMT) systems have previously been used for the task of grammatical error correction (GEC) to achieve state-of-the-art accuracy. The superiority of SMT systems comes from their ability to learn text transformations from erroneous to corrected text, without explicitly modeling error types. However, phrase-based SMT systems suffer from limitations of discrete word representation, linear mapping, and lack of global context. In this paper, we address these limitations by using two different yet complementary neural network models, namely a neural network global lexicon model and a neural network joint model. These neural networks can generalize better by using continuous space representation of words and learn non-linear mappings. Moreover, they can leverage contextual information from the source sentence more effectively. By adding these two components, we achieve statistically significant improvement in accuracy for grammatical error correction over a state-of-the-art GEC system.