Lei Shu

Zixuan Ke, Haowei Lin, Yijia Shao et al. · deepmind, pku

Recent work on applying large language models (LMs) achieves impressive performance in many NLP applications. Adapting or posttraining an LM using an unlabeled domain corpus can produce even better performance for end-tasks in the domain. This paper proposes the problem of continually extending an LM by incrementally post-train the LM with a sequence of unlabeled domain corpora to expand its knowledge without forgetting its previous skills. The goal is to improve the few-shot end-task learning in these domains. The resulting system is called CPT (Continual PostTraining), which to our knowledge, is the first continual post-training system. Experimental results verify its effectiveness.

Zixuan Ke, Yijia Shao, Haowei Lin et al. · deepmind, pku

Domain-adaptive pre-training (or DA-training for short), also known as post-training, aims to train a pre-trained general-purpose language model (LM) using an unlabeled corpus of a particular domain to adapt the LM so that end-tasks in the domain can give improved performances. However, existing DA-training methods are in some sense blind as they do not explicitly identify what knowledge in the LM should be preserved and what should be changed by the domain corpus. This paper shows that the existing methods are suboptimal and proposes a novel method to perform a more informed adaptation of the knowledge in the LM by (1) soft-masking the attention heads based on their importance to best preserve the general knowledge in the LM and (2) contrasting the representations of the general and the full (both general and domain knowledge) to learn an integrated representation with both general and domain-specific knowledge. Experimental results will demonstrate the effectiveness of the proposed approach.

Yun Zhu, Yinxiao Liu, Felix Stahlberg et al. · deepmind

Large Language Models (LLMs) have demonstrated impressive capabilities for text rewriting. Nonetheless, the large sizes of these models make them impractical for on-device inference, which would otherwise allow for enhanced privacy and economical inference. Creating a smaller yet potent language model for text rewriting presents a formidable challenge because it requires balancing the need for a small size with the need to retain the emergent capabilities of the LLM, that requires costly data collection. To address the above challenge, we introduce a new instruction tuning approach for building a mobile-centric text rewriting model. Our strategies enable the generation of high quality training data without any human labeling. In addition, we propose a heuristic reinforcement learning framework which substantially enhances performance without requiring preference data. To further bridge the performance gap with the larger server-side model, we propose an effective approach that combines the mobile rewrite agent with the server model using a cascade. To tailor the text rewriting tasks to mobile scenarios, we introduce MessageRewriteEval, a benchmark that focuses on text rewriting for messages through natural language instructions. Through empirical experiments, we demonstrate that our on-device model surpasses the current state-of-the-art LLMs in text rewriting while maintaining a significantly reduced model size. Notably, we show that our proposed cascading approach improves model performance.

Yipeng Gao, Lei Shu, Genzhi Ye et al.

Procedural 3D modeling through code is emerging as a versatile paradigm, offering deterministic, engine-ready, and precisely editable assets that neural 3D generators inherently lack. Authoring such procedural content, however, demands deep expertise in 3D software APIs, parametric design, and code-level geometric reasoning. In this paper, we propose 3DCodeBench, a systematic benchmark for evaluating vision-language model (VLM) agents for procedural 3D generation in 3D modeling software. Specifically, 3DCodeBench evaluates how effectively 12 advanced VLMs can serve as procedural 3D modelers by translating text and image references into procedural code for 3D modeling software. Recognizing that automated metrics may not fully capture the perceptual quality of 3D shapes, we build 3DCodeArena, a ranking platform based on pairwise human preferences over generated 3D outputs. From extensive evaluations and results, we observe that: (1) Failures mostly arise from API mismatches, while successful renders still suffer from disconnected or floating 3D geometric components. (2) Test-time scaling, such as higher thinking budgets and multi-turn refinement, improves performance overall. Our findings highlight a critical need for high-quality procedural coding data to advance commercial VLMs. Furthermore, effective procedural 3D modeling requires a robust execution environment that provides high-fidelity feedback for iterative refinement. We release 3DCodeBench, including the curated large-scale dataset of multimodal (text/image) prompts, procedural code, 3D object triplets, evaluation protocol, and the public 3DCodeArena platform as a foundational toolkit for exploring VLM-based procedural 3D modelers.

Maria Wang, Srinivas Sunkara, Gilles Baechler et al.

The rise of powerful multimodal LLMs has enhanced the viability of building web agents which can, with increasing levels of autonomy, assist users to retrieve information and complete tasks on various human-computer interfaces. It is hence necessary to build challenging benchmarks that span a wide-variety of use cases reflecting real-world usage. In this work, we present WebQuest, a multi-page question-answering dataset that requires reasoning across multiple related web pages. In contrast to existing UI benchmarks that focus on multi-step web navigation and task completion, our dataset evaluates information extraction, multimodal retrieval and composition of information from many web pages. WebQuest includes three question categories: single-screen QA, multi-screen QA, and QA based on navigation traces. We evaluate leading proprietary multimodal models like GPT-4V, Gemini Flash, Claude 3, and open source models like InstructBLIP, PaliGemma on our dataset, revealing a significant gap between single-screen and multi-screen reasoning. Finally, we investigate inference time techniques like Chain-of-Thought prompting to improve model capabilities on multi-screen reasoning.

Sepideh Esmaeilpour, Lei Shu, Bing Liu

The primary assumption of conventional supervised learning or classification is that the test samples are drawn from the same distribution as the training samples, which is called closed set learning or classification. In many practical scenarios, this is not the case because there are unknowns or unseen class samples in the test data, which is called the open set scenario, and the unknowns need to be detected. This problem is referred to as the open set recognition problem and is important in safety-critical applications. We propose to detect unknowns (or unseen class samples) through learning pairwise similarities. The proposed method works in two steps. It first learns a closed set classifier using the seen classes that have appeared in training and then learns how to compare seen classes with pseudo-unseen (automatically generated unseen class samples). The pseudo-unseen generation is carried out by performing distribution shifting augmentations on the seen or training samples. We call our method OPG (Open set recognition based on Pseudo unseen data Generation). The experimental evaluation shows that the learned similarity-based features can successfully distinguish seen from unseen in benchmark datasets for open set recognition.

Yun Zhu, Nevan Wichers, Chu-Cheng Lin et al.

Parameter Efficient Tuning has been an prominent approach to adapt the Large Language Model to downstream tasks. Most previous works considers adding the dense trainable parameters, where all parameters are used to adapt certain task. We found this less effective empirically using the example of LoRA that introducing more trainable parameters does not help. Motivated by this we investigate the importance of leveraging "sparse" computation and propose SiRA: sparse mixture of low rank adaption. SiRA leverages the Sparse Mixture of Expert(SMoE) to boost the performance of LoRA. Specifically it enforces the top $k$ experts routing with a capacity limit restricting the maximum number of tokens each expert can process. We propose a novel and simple expert dropout on top of gating network to reduce the over-fitting issue. Through extensive experiments, we verify SiRA performs better than LoRA and other mixture of expert approaches across different single tasks and multitask settings.

Liangchen Luo, Zi Lin, Yinxiao Liu et al.

Critical thinking is essential for rational decision-making and problem-solving. This skill hinges on the ability to provide precise and reasoned critiques and is a hallmark of human intelligence. In the era of large language models (LLMs), this study explores the ability of LLMs to deliver accurate critiques across various tasks. We are interested in this topic as a capable critic model could not only serve as a reliable evaluator, but also as a source of supervised signals for model tuning. Particularly, if a model can self-critique, it has the potential for autonomous self-improvement. To examine this, we introduce a unified evaluation framework for assessing the critique abilities of LLMs. We develop a benchmark called CriticBench, which comprises 3K high-quality natural language queries and corresponding model responses; and annotate the correctness of these responses. The benchmark cover tasks such as math problem-solving, code completion, and question answering. We evaluate multiple LLMs on the collected dataset and our analysis reveals several noteworthy insights: (1) Critique is generally challenging for most LLMs, and this capability often emerges only when models are sufficiently large. (2) In particular, self-critique is especially difficult. Even top-performing LLMs struggle to achieve satisfactory performance. (3) Models tend to have lower critique accuracy on problems where they are most uncertain. To this end, we introduce a simple yet effective baseline named self-check, which leverages self-critique to improve task performance for various models. We hope this study serves as an initial exploration into understanding the critique abilities of LLMs, and aims to inform future research, including the development of more proficient critic models and the application of critiques across diverse tasks.

Lei Shu, Dong Zhao, Jianli Chen et al.

Residential energy retrofit initiation is often stalled by an expertise gap, where homeowners lack the technical literacy required for structured building energy assessments and are thereby trapped in low-information environments with fragmented sources. To bridge this gap, this study reports a domain-specific large language model (LLM) designed to catalyze informed decision-making based solely on homeowner-accessible, natural-language descriptions, e.g., building age, size, and location. The model is created using the parameter-efficient low-rank adaption (LoRA) fine-tuning approach on a massive corpus grounded in physics-based energy simulations and techno-economic calculations from 536,416 U.S. residential building prototypes. Nine major retrofit categories are evaluated, including envelope upgrades, HVAC systems, and renewable energy installations. Validations against physics-grounded benchmarks show that the LLM consistently identifies high-quality retrofit options, achieving top-3 hit rates of 98.9% for maximum CO2 reduction and 93.3% for the shortest discounted payback year. Moreover, the model exhibits strong robustness under incomplete input conditions, maintaining stable performance even when basic dwelling descriptions are only 60% partially specified. By significantly lowering the information activation energy for non-expert users while maintaining the scientific rigor, this physics-based AI model offers a scalable pathway for parallelized, user-centered decision making, accelerating cumulative energy savings and emission reductions across community and national scales.

Lei Shu, Vinicius Azevedo, Barbara Solenthaler et al.

The accurate representation of fine-detailed cloth wrinkles poses significant challenges in computer graphics. The inherently non-uniform structure of cloth wrinkles mandates the employment of intricate discretization strategies, which are frequently characterized by high computational demands and complex methodologies. Addressing this, the research introduced in this paper elucidates a novel anisotropic cloth regression technique that capitalizes on the potential of implicit neural representations of surfaces. Our first core contribution is an innovative mesh-free sampling approach, crafted to reduce the reliance on traditional mesh structures, thereby offering greater flexibility and accuracy in capturing fine cloth details. Our second contribution is a novel adversarial training scheme, which is designed meticulously to strike a harmonious balance between the sampling and simulation objectives. The adversarial approach ensures that the wrinkles are represented with high fidelity, while also maintaining computational efficiency. Our results showcase through various cloth-object interaction scenarios that our method, given the same memory constraints, consistently surpasses traditional discrete representations, particularly when modelling highly-detailed localized wrinkles.

Lei Shu, Nevan Wichers, Liangchen Luo et al.

Evaluating natural language systems poses significant challenges, particularly in the realms of natural language understanding and high-level reasoning. In this paper, we introduce 'Fusion-Eval', an innovative approach that leverages Large Language Models (LLMs) to integrate insights from various assistant evaluators. The LLM is given the example to evaluate along with scores from the assistant evaluators. Each of these evaluators specializes in assessing distinct aspects of responses. Fusion-Eval achieves a 0.962 system-level Kendall-Tau correlation with humans on SummEval and a 0.744 turn-level Spearman correlation on TopicalChat, which is significantly higher than baseline methods. These results highlight Fusion-Eval's significant potential in the realm of natural language system evaluation.

Shoubin Yu, Lei Shu, Antoine Yang et al.

Multimodal AI agents are increasingly automating complex real-world workflows that involve online web execution. However, current web-agent benchmarks suffer from a critical limitation: they focus entirely on web-based interaction and perception, lacking grounding in the user's real-world physical surroundings. This limitation prevents evaluation in crucial scenarios, such as when an agent must use egocentric visual perception (e.g., via AR glasses) to recognize an object in the user's surroundings and then complete a related task online. To address this gap, we introduce Ego2Web, the first benchmark designed to bridge egocentric video perception and web agent execution. Ego2Web pairs real-world first-person video recordings with web tasks that require visual understanding, web task planning, and interaction in an online environment for successful completion. We utilize an automatic data-generation pipeline combined with human verification and refinement to curate well-constructed, high-quality video-task pairs across diverse web task types, including e-commerce, media retrieval, knowledge lookup, etc. To facilitate accurate and scalable evaluation for our benchmark, we also develop a novel LLM-as-a-Judge automatic evaluation method, Ego2WebJudge, which achieves approximately 84% agreement with human judgment, substantially higher than existing evaluation methods. Experiments with diverse SoTA agents on our Ego2Web show that their performance is weak, with substantial headroom across all task categories. We also conduct a comprehensive ablation study on task design, highlighting the necessity of accurate video understanding in the proposed task and the limitations of current agents. We hope Ego2Web can be a critical new resource for developing truly capable AI assistants that can seamlessly see, understand, and act across the physical and digital worlds.

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Tao Zhu, Qi Yu, Xinru Dong et al.

Weakly-supervised video anomaly detection (WS-VAD) using Multiple Instance Learning (MIL) suffers from label ambiguity, hindering discriminative feature learning. We propose ProDisc-VAD, an efficient framework tackling this via two synergistic components. The Prototype Interaction Layer (PIL) provides controlled normality modeling using a small set of learnable prototypes, establishing a robust baseline without being overwhelmed by dominant normal data. The Pseudo-Instance Discriminative Enhancement (PIDE) loss boosts separability by applying targeted contrastive learning exclusively to the most reliable extreme-scoring instances (highest/lowest scores). ProDisc-VAD achieves strong AUCs (97.98% ShanghaiTech, 87.12% UCF-Crime) using only 0.4M parameters, over 800x fewer than recent ViT-based methods like VadCLIP. Code is available at https://github.com/modadundun/ProDisc-VAD.

Lei Shu, Liangchen Luo, Jayakumar Hoskere et al.

Large Language Models (LLMs) have demonstrated impressive capabilities in creative tasks such as storytelling and E-mail generation. However, as LLMs are primarily trained on final text results rather than intermediate revisions, it might be challenging for them to perform text rewriting tasks. Most studies in the rewriting tasks focus on a particular transformation type within the boundaries of single sentences. In this work, we develop new strategies for instruction tuning and reinforcement learning to better align LLMs for cross-sentence rewriting tasks using diverse wording and structures expressed through natural languages including 1) generating rewriting instruction data from Wiki edits and public corpus through instruction generation and chain-of-thought prompting; 2) collecting comparison data for reward model training through a new ranking function. To facilitate this research, we introduce OpenRewriteEval, a novel benchmark covers a wide variety of rewriting types expressed through natural language instructions. Our results show significant improvements over a variety of baselines. The public repository is available on GitHub under Google Research (https://github.com/google-research/google-research/tree/master/rewritelm).

Hu Xu, Lei Shu, Philip S. Yu et al.

This paper analyzes the pre-trained hidden representations learned from reviews on BERT for tasks in aspect-based sentiment analysis (ABSA). Our work is motivated by the recent progress in BERT-based language models for ABSA. However, it is not clear how the general proxy task of (masked) language model trained on unlabeled corpus without annotations of aspects or opinions can provide important features for downstream tasks in ABSA. By leveraging the annotated datasets in ABSA, we investigate both the attentions and the learned representations of BERT pre-trained on reviews. We found that BERT uses very few self-attention heads to encode context words (such as prepositions or pronouns that indicating an aspect) and opinion words for an aspect. Most features in the representation of an aspect are dedicated to the fine-grained semantics of the domain (or product category) and the aspect itself, instead of carrying summarized opinions from its context. We hope this investigation can help future research in improving self-supervised learning, unsupervised learning and fine-tuning for ABSA. The pre-trained model and code can be found at https://github.com/howardhsu/BERT-for-RRC-ABSA.

Hu Xu, Bing Liu, Lei Shu et al.

Aspect-based sentiment classification (ASC) is an important task in fine-grained sentiment analysis.~Deep supervised ASC approaches typically model this task as a pair-wise classification task that takes an aspect and a sentence containing the aspect and outputs the polarity of the aspect in that sentence. However, we discovered that many existing approaches fail to learn an effective ASC classifier but more like a sentence-level sentiment classifier because they have difficulty to handle sentences with different polarities for different aspects.~This paper first demonstrates this problem using several state-of-the-art ASC models. It then proposes a novel and general adaptive re-weighting (ARW) scheme to adjust the training to dramatically improve ASC for such complex sentences. Experimental results show that the proposed framework is effective \footnote{The dataset and code are available at \url{https://github.com/howardhsu/ASC_failure}.}.

Lei Shu, Hu Xu, Bing Liu et al.

Dialogue management (DM) plays a key role in the quality of the interaction with the user in a task-oriented dialogue system. In most existing approaches, the agent predicts only one DM policy action per turn. This significantly limits the expressive power of the conversational agent and introduces unwanted turns of interactions that may challenge users' patience. Longer conversations also lead to more errors and the system needs to be more robust to handle them. In this paper, we compare the performance of several models on the task of predicting multiple acts for each turn. A novel policy model is proposed based on a recurrent cell called gated Continue-Act-Slots (gCAS) that overcomes the limitations of the existing models. Experimental results show that gCAS outperforms other approaches. The code is available at https://leishu02.github.io/

Lei Shu, Piero Molino, Mahdi Namazifar et al.

This paper proposes a novel end-to-end architecture for task-oriented dialogue systems. It is based on a simple and practical yet very effective sequence-to-sequence approach, where language understanding and state tracking tasks are modeled jointly with a structured copy-augmented sequential decoder and a multi-label decoder for each slot. The policy engine and language generation tasks are modeled jointly following that. The copy-augmented sequential decoder deals with new or unknown values in the conversation, while the multi-label decoder combined with the sequential decoder ensures the explicit assignment of values to slots. On the generation part, slot binary classifiers are used to improve performance. This architecture is scalable to real-world scenarios and is shown through an empirical evaluation to achieve state-of-the-art performance on both the Cambridge Restaurant dataset and the Stanford in-car assistant dataset\footnote{The code is available at \url{https://github.com/uber-research/FSDM}}

Hu Xu, Bing Liu, Lei Shu et al.

Inspired by conversational reading comprehension (CRC), this paper studies a novel task of leveraging reviews as a source to build an agent that can answer multi-turn questions from potential consumers of online businesses. We first build a review CRC dataset and then propose a novel task-aware pre-tuning step running between language model (e.g., BERT) pre-training and domain-specific fine-tuning. The proposed pre-tuning requires no data annotation, but can greatly enhance the performance on our end task. Experimental results show that the proposed approach is highly effective and has competitive performance as the supervised approach. The dataset is available at \url{https://github.com/howardhsu/RCRC}

Meng Cao, Lei Shu, Lei Yu et al.

Reinforcement learning (RL) can align language models with non-differentiable reward signals, such as human preferences. However, a major challenge arises from the sparsity of these reward signals - typically, there is only a single reward for an entire output. This sparsity of rewards can lead to inefficient and unstable learning. To address this challenge, our paper introduces an novel framework that utilizes the critique capability of Large Language Models (LLMs) to produce intermediate-step rewards during RL training. Our method involves coupling a policy model with a critic language model, which is responsible for providing comprehensive feedback of each part of the output. This feedback is then translated into token or span-level rewards that can be used to guide the RL training process. We investigate this approach under two different settings: one where the policy model is smaller and is paired with a more powerful critic model, and another where a single language model fulfills both roles. We assess our approach on three text generation tasks: sentiment control, language model detoxification, and summarization. Experimental results show that incorporating artificial intrinsic rewards significantly improve both sample efficiency and the overall performance of the policy model, supported by both automatic and human evaluation.

Yufei Li, John Nham, Ganesh Jawahar et al.

Generic text rewriting is a prevalent large language model (LLM) application that covers diverse real-world tasks, such as style transfer, fact correction, and email editing. These tasks vary in rewriting objectives (e.g., factual consistency vs. semantic preservation), making it challenging to develop a unified model that excels across all dimensions. Existing methods often specialize in either a single task or a specific objective, limiting their generalizability. In this work, we introduce a generic model proficient in factuality, stylistic, and conversational rewriting tasks. To simulate real-world user rewrite requests, we construct a conversational rewrite dataset, ChatRewrite, that presents ``natural''-sounding instructions, from raw emails using LLMs. Combined with other popular rewrite datasets, including LongFact for the factuality rewrite task and RewriteLM for the stylistic rewrite task, this forms a broad benchmark for training and evaluating generic rewrite models. To align with task-specific objectives, we propose Dr Genre, a Decoupled-reward learning framework for Generic rewriting, that utilizes objective-oriented reward models with a task-specific weighting. Evaluation shows that \approach delivers higher-quality rewrites across all targeted tasks, improving objectives including instruction following (agreement), internal consistency (coherence), and minimal unnecessary edits (conciseness).

Lei Shu, Dong Zhao

Conventional approaches to building energy retrofit decision making suffer from limited generalizability and low interpretability, hindering adoption in diverse residential contexts. With the growth of Smart and Connected Communities, generative AI, especially large language models (LLMs), may help by processing contextual information and producing practitioner readable recommendations. We evaluate seven LLMs (ChatGPT, DeepSeek, Gemini, Grok, Llama, and Claude) on residential retrofit decisions under two objectives: maximizing CO2 reduction (technical) and minimizing payback period (sociotechnical). Performance is assessed on four dimensions: accuracy, consistency, sensitivity, and reasoning, using a dataset of 400 homes across 49 US states. LLMs generate effective recommendations in many cases, reaching up to 54.5 percent top 1 match and 92.8 percent within top 5 without fine tuning. Performance is stronger for the technical objective, while sociotechnical decisions are limited by economic trade offs and local context. Agreement across models is low, and higher performing models tend to diverge from others. LLMs are sensitive to location and building geometry but less sensitive to technology and occupant behavior. Most models show step by step, engineering style reasoning, but it is often simplified and lacks deeper contextual awareness. Overall, LLMs are promising assistants for energy retrofit decision making, but improvements in accuracy, consistency, and context handling are needed for reliable practice.

Jiannan Xiang, Yun Zhu, Lei Shu et al.

Developing and testing user interfaces (UIs) and training AI agents to interact with them are challenging due to the dynamic and diverse nature of real-world mobile environments. Existing methods often rely on cumbersome physical devices or limited static analysis of screenshots, which hinders scalable testing and the development of intelligent UI agents. We introduce UISim, a novel image-based UI simulator that offers a dynamic and interactive platform for exploring mobile phone environments purely from screen images. Our system employs a two-stage method: given an initial phone screen image and a user action, it first predicts the abstract layout of the next UI state, then synthesizes a new, visually consistent image based on this predicted layout. This approach enables the realistic simulation of UI transitions. UISim provides immediate practical benefits for UI testing, rapid prototyping, and synthetic data generation. Furthermore, its interactive capabilities pave the way for advanced applications, such as UI navigation task planning for AI agents. Our experimental results show that UISim outperforms end-to-end UI generation baselines in generating realistic and coherent subsequent UI states, highlighting its fidelity and potential to streamline UI development and enhance AI agent training.

Liangchen Luo, Yinxiao Liu, Rosanne Liu et al.

Complex multi-step reasoning tasks, such as solving mathematical problems or generating code, remain a significant hurdle for even the most advanced large language models (LLMs). Verifying LLM outputs with an Outcome Reward Model (ORM) is a standard inference-time technique aimed at enhancing the reasoning performance of LLMs. However, this still proves insufficient for reasoning tasks with a lengthy or multi-hop reasoning chain, where the intermediate outcomes are neither properly rewarded nor penalized. Process supervision addresses this limitation by assigning intermediate rewards during the reasoning process. To date, the methods used to collect process supervision data have relied on either human annotation or per-step Monte Carlo estimation, both prohibitively expensive to scale, thus hindering the broad application of this technique. In response to this challenge, we propose a novel divide-and-conquer style Monte Carlo Tree Search (MCTS) algorithm named \textit{OmegaPRM} for the efficient collection of high-quality process supervision data. This algorithm swiftly identifies the first error in the Chain of Thought (CoT) with binary search and balances the positive and negative examples, thereby ensuring both efficiency and quality. As a result, we are able to collect over 1.5 million process supervision annotations to train Process Reward Models (PRMs). This fully automated process supervision alongside the weighted self-consistency algorithm is able to enhance LLMs' math reasoning performances. We improved the success rates of the instruction-tuned Gemini Pro model from 51\% to 69.4\% on MATH500 and from 86.4\% to 93.6\% on GSM8K. Similarly, we boosted the success rates of Gemma2 27B from 42.3\% to 58.2\% on MATH500 and from 74.0\% to 92.2\% on GSM8K. The entire process operates without any human intervention or supervision, making our method both financially and ...

Deng Cai, Elman Mansimov, Yi-An Lai et al.

Recent advance in deep learning has led to the rapid adoption of machine learning-based NLP models in a wide range of applications. Despite the continuous gain in accuracy, backward compatibility is also an important aspect for industrial applications, yet it received little research attention. Backward compatibility requires that the new model does not regress on cases that were correctly handled by its predecessor. This work studies model update regression in structured prediction tasks. We choose syntactic dependency parsing and conversational semantic parsing as representative examples of structured prediction tasks in NLP. First, we measure and analyze model update regression in different model update settings. Next, we explore and benchmark existing techniques for reducing model update regression including model ensemble and knowledge distillation. We further propose a simple and effective method, Backward-Congruent Re-ranking (BCR), by taking into account the characteristics of structured prediction. Experiments show that BCR can better mitigate model update regression than model ensemble and knowledge distillation approaches.

Lei Shu, Hu Xu, Bing Liu et al.

Aspect-based sentiment analysis (ABSA) typically requires in-domain annotated data for supervised training/fine-tuning. It is a big challenge to scale ABSA to a large number of new domains. This paper aims to train a unified model that can perform zero-shot ABSA without using any annotated data for a new domain. We propose a method called contrastive post-training on review Natural Language Inference (CORN). Later ABSA tasks can be cast into NLI for zero-shot transfer. We evaluate CORN on ABSA tasks, ranging from aspect extraction (AE), aspect sentiment classification (ASC), to end-to-end aspect-based sentiment analysis (E2E ABSA), which show ABSA can be conducted without any human annotated ABSA data.

Zixuan Ke, Bing Liu, Hao Wang et al.

This paper studies continual learning (CL) for sentiment classification (SC). In this setting, the CL system learns a sequence of SC tasks incrementally in a neural network, where each task builds a classifier to classify the sentiment of reviews of a particular product category or domain. Two natural questions are: Can the system transfer the knowledge learned in the past from the previous tasks to the new task to help it learn a better model for the new task? And, can old models for previous tasks be improved in the process as well? This paper proposes a novel technique called KAN to achieve these objectives. KAN can markedly improve the SC accuracy of both the new task and the old tasks via forward and backward knowledge transfer. The effectiveness of KAN is demonstrated through extensive experiments.

Zixuan Ke, Bing Liu, Hu Xu et al.

This paper studies continual learning (CL) of a sequence of aspect sentiment classification(ASC) tasks in a particular CL setting called domain incremental learning (DIL). Each task is from a different domain or product. The DIL setting is particularly suited to ASC because in testing the system needs not know the task/domain to which the test data belongs. To our knowledge, this setting has not been studied before for ASC. This paper proposes a novel model called CLASSIC. The key novelty is a contrastive continual learning method that enables both knowledge transfer across tasks and knowledge distillation from old tasks to the new task, which eliminates the need for task ids in testing. Experimental results show the high effectiveness of CLASSIC.

Zixuan Ke, Bing Liu, Nianzu Ma et al.

Continual learning (CL) learns a sequence of tasks incrementally with the goal of achieving two main objectives: overcoming catastrophic forgetting (CF) and encouraging knowledge transfer (KT) across tasks. However, most existing techniques focus only on overcoming CF and have no mechanism to encourage KT, and thus do not do well in KT. Although several papers have tried to deal with both CF and KT, our experiments show that they suffer from serious CF when the tasks do not have much shared knowledge. Another observation is that most current CL methods do not use pre-trained models, but it has been shown that such models can significantly improve the end task performance. For example, in natural language processing, fine-tuning a BERT-like pre-trained language model is one of the most effective approaches. However, for CL, this approach suffers from serious CF. An interesting question is how to make the best use of pre-trained models for CL. This paper proposes a novel model called CTR to solve these problems. Our experimental results demonstrate the effectiveness of CTR

Yixuan Su, Fangyu Liu, Zaiqiao Meng et al.

Masked language models (MLMs) such as BERT and RoBERTa have revolutionized the field of Natural Language Understanding in the past few years. However, existing pre-trained MLMs often output an anisotropic distribution of token representations that occupies a narrow subset of the entire representation space. Such token representations are not ideal, especially for tasks that demand discriminative semantic meanings of distinct tokens. In this work, we propose TaCL (Token-aware Contrastive Learning), a novel continual pre-training approach that encourages BERT to learn an isotropic and discriminative distribution of token representations. TaCL is fully unsupervised and requires no additional data. We extensively test our approach on a wide range of English and Chinese benchmarks. The results show that TaCL brings consistent and notable improvements over the original BERT model. Furthermore, we conduct detailed analysis to reveal the merits and inner-workings of our approach.

Chao Qi, Junfeng Gao, Simon Pearson et al.

Tea chrysanthemum detection at its flowering stage is one of the key components for selective chrysanthemum harvesting robot development. However, it is a challenge to detect flowering chrysanthemums under unstructured field environments given the variations on illumination, occlusion and object scale. In this context, we propose a highly fused and lightweight deep learning architecture based on YOLO for tea chrysanthemum detection (TC-YOLO). First, in the backbone component and neck component, the method uses the Cross-Stage Partially Dense Network (CSPDenseNet) as the main network, and embeds custom feature fusion modules to guide the gradient flow. In the final head component, the method combines the recursive feature pyramid (RFP) multiscale fusion reflow structure and the Atrous Spatial Pyramid Pool (ASPP) module with cavity convolution to achieve the detection task. The resulting model was tested on 300 field images, showing that under the NVIDIA Tesla P100 GPU environment, if the inference speed is 47.23 FPS for each image (416 * 416), TC-YOLO can achieve the average precision (AP) of 92.49% on our own tea chrysanthemum dataset. In addition, this method (13.6M) can be deployed on a single mobile GPU, and it could be further developed as a perception system for a selective chrysanthemum harvesting robot in the future.

Yixuan Su, Lei Shu, Elman Mansimov et al.

Pre-trained language models have been recently shown to benefit task-oriented dialogue (TOD) systems. Despite their success, existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub-tasks and greater data annotation overhead. In this study, we present PPTOD, a unified plug-and-play model for task-oriented dialogue. In addition, we introduce a new dialogue multi-task pre-training strategy that allows the model to learn the primary TOD task completion skills from heterogeneous dialog corpora. We extensively test our model on three benchmark TOD tasks, including end-to-end dialogue modelling, dialogue state tracking, and intent classification. Experimental results show that PPTOD achieves new state of the art on all evaluated tasks in both high-resource and low-resource scenarios. Furthermore, comparisons against previous SOTA methods show that the responses generated by PPTOD are more factually correct and semantically coherent as judged by human annotators.

Sepideh Esmaeilpour, Bing Liu, Eric Robertson et al.

In an out-of-distribution (OOD) detection problem, samples of known classes(also called in-distribution classes) are used to train a special classifier. In testing, the classifier can (1) classify the test samples of known classes to their respective classes and also (2) detect samples that do not belong to any of the known classes (i.e., they belong to some unknown or OOD classes). This paper studies the problem of zero-shot out-of-distribution(OOD) detection, which still performs the same two tasks in testing but has no training except using the given known class names. This paper proposes a novel yet simple method (called ZOC) to solve the problem. ZOC builds on top of the recent advances in zero-shot classification through multi-modal representation learning. It first extends the pre-trained language-vision model CLIP by training a text-based image description generator on top of CLIP. In testing, it uses the extended model to generate candidate unknown class names for each test sample and computes a confidence score based on both the known class names and candidate unknown class names for zero-shot OOD detection. Experimental results on 5 benchmark datasets for OOD detection demonstrate that ZOC outperforms the baselines by a large margin.

Lei Shu, Alexandros Papangelis, Yi-Chia Wang et al.

This work introduces Focused-Variation Network (FVN), a novel model to control language generation. The main problems in previous controlled language generation models range from the difficulty of generating text according to the given attributes, to the lack of diversity of the generated texts. FVN addresses these issues by learning disjoint discrete latent spaces for each attribute inside codebooks, which allows for both controllability and diversity, while at the same time generating fluent text. We evaluate FVN on two text generation datasets with annotated content and style, and show state-of-the-art performance as assessed by automatic and human evaluations.

Hu Xu, Bing Liu, Lei Shu et al.

This paper focuses on learning domain-oriented language models driven by end tasks, which aims to combine the worlds of both general-purpose language models (such as ELMo and BERT) and domain-specific language understanding. We propose DomBERT, an extension of BERT to learn from both in-domain corpus and relevant domain corpora. This helps in learning domain language models with low-resources. Experiments are conducted on an assortment of tasks in aspect-based sentiment analysis, demonstrating promising results.

Lei Shu, Hu Xu, Bing Liu

One key task of fine-grained sentiment analysis on reviews is to extract aspects or features that users have expressed opinions on. This paper focuses on supervised aspect extraction using a modified CNN called controlled CNN (Ctrl). The modified CNN has two types of control modules. Through asynchronous parameter updating, it prevents over-fitting and boosts CNN's performance significantly. This model achieves state-of-the-art results on standard aspect extraction datasets. To the best of our knowledge, this is the first paper to apply control modules to aspect extraction.

Hu Xu, Bing Liu, Lei Shu et al.

Question-answering plays an important role in e-commerce as it allows potential customers to actively seek crucial information about products or services to help their purchase decision making. Inspired by the recent success of machine reading comprehension (MRC) on formal documents, this paper explores the potential of turning customer reviews into a large source of knowledge that can be exploited to answer user questions.~We call this problem Review Reading Comprehension (RRC). To the best of our knowledge, no existing work has been done on RRC. In this work, we first build an RRC dataset called ReviewRC based on a popular benchmark for aspect-based sentiment analysis. Since ReviewRC has limited training examples for RRC (and also for aspect-based sentiment analysis), we then explore a novel post-training approach on the popular language model BERT to enhance the performance of fine-tuning of BERT for RRC. To show the generality of the approach, the proposed post-training is also applied to some other review-based tasks such as aspect extraction and aspect sentiment classification in aspect-based sentiment analysis. Experimental results demonstrate that the proposed post-training is highly effective. The datasets and code are available at https://www.cs.uic.edu/~hxu/.

Hu Xu, Bing Liu, Lei Shu et al.

Classic supervised learning makes the closed-world assumption, meaning that classes seen in testing must have been seen in training. However, in the dynamic world, new or unseen class examples may appear constantly. A model working in such an environment must be able to reject unseen classes (not seen or used in training). If enough data is collected for the unseen classes, the system should incrementally learn to accept/classify them. This learning paradigm is called open-world learning (OWL). Existing OWL methods all need some form of re-training to accept or include the new classes in the overall model. In this paper, we propose a meta-learning approach to the problem. Its key novelty is that it only needs to train a meta-classifier, which can then continually accept new classes when they have enough labeled data for the meta-classifier to use, and also detect/reject future unseen classes. No re-training of the meta-classifier or a new overall classifier covering all old and new classes is needed. In testing, the method only uses the examples of the seen classes (including the newly added classes) on-the-fly for classification and rejection. Experimental results demonstrate the effectiveness of the new approach.

Hu Xu, Bing Liu, Lei Shu et al.

Learning high-quality domain word embeddings is important for achieving good performance in many NLP tasks. General-purpose embeddings trained on large-scale corpora are often sub-optimal for domain-specific applications. However, domain-specific tasks often do not have large in-domain corpora for training high-quality domain embeddings. In this paper, we propose a novel lifelong learning setting for domain embedding. That is, when performing the new domain embedding, the system has seen many past domains, and it tries to expand the new in-domain corpus by exploiting the corpora from the past domains via meta-learning. The proposed meta-learner characterizes the similarities of the contexts of the same word in many domain corpora, which helps retrieve relevant data from the past domains to expand the new domain corpus. Experimental results show that domain embeddings produced from such a process improve the performance of the downstream tasks.

Hu Xu, Bing Liu, Lei Shu et al.

One key task of fine-grained sentiment analysis of product reviews is to extract product aspects or features that users have expressed opinions on. This paper focuses on supervised aspect extraction using deep learning. Unlike other highly sophisticated supervised deep learning models, this paper proposes a novel and yet simple CNN model employing two types of pre-trained embeddings for aspect extraction: general-purpose embeddings and domain-specific embeddings. Without using any additional supervision, this model achieves surprisingly good results, outperforming state-of-the-art sophisticated existing methods. To our knowledge, this paper is the first to report such double embeddings based CNN model for aspect extraction and achieve very good results.

Zhaopeng Tu, Yong Jiang, Xiaojiang Liu et al.

We study the problem of stock related question answering (StockQA): automatically generating answers to stock related questions, just like professional stock analysts providing action recommendations to stocks upon user's requests. StockQA is quite different from previous QA tasks since (1) the answers in StockQA are natural language sentences (rather than entities or values) and due to the dynamic nature of StockQA, it is scarcely possible to get reasonable answers in an extractive way from the training data; and (2) StockQA requires properly analyzing the relationship between keywords in QA pair and the numerical features of a stock. We propose to address the problem with a memory-augmented encoder-decoder architecture, and integrate different mechanisms of number understanding and generation, which is a critical component of StockQA. We build a large-scale dataset containing over 180K StockQA instances, based on which various technique combinations are extensively studied and compared. Experimental results show that a hybrid word-character model with separate character components for number processing, achieves the best performance. By analyzing the results, we found that 44.8% of answers generated by our best model still suffer from the generic answer problem, which can be alleviated by a straightforward hybrid retrieval-generation model.

Lei Shu, Hu Xu, Bing Liu

This paper concerns open-world classification, where the classifier not only needs to classify test examples into seen classes that have appeared in training but also reject examples from unseen or novel classes that have not appeared in training. Specifically, this paper focuses on discovering the hidden unseen classes of the rejected examples. Clearly, without prior knowledge this is difficult. However, we do have the data from the seen training classes, which can tell us what kind of similarity/difference is expected for examples from the same class or from different classes. It is reasonable to assume that this knowledge can be transferred to the rejected examples and used to discover the hidden unseen classes in them. This paper aims to solve this problem. It first proposes a joint open classification model with a sub-model for classifying whether a pair of examples belongs to the same or different classes. This sub-model can serve as a distance function for clustering to discover the hidden classes of the rejected examples. Experimental results show that the proposed model is highly promising.

Hu Xu, Sihong Xie, Lei Shu et al.

Functionality is of utmost importance to customers when they purchase products. However, it is unclear to customers whether a product can really satisfy their needs on functions. Further, missing functions may be intentionally hidden by the manufacturers or the sellers. As a result, a customer needs to spend a fair amount of time before purchasing or just purchase the product on his/her own risk. In this paper, we first identify a novel QA corpus that is dense on product functionality information \footnote{The annotated corpus can be found at \url{https://www.cs.uic.edu/~hxu/}.}. We then design a neural network called Semi-supervised Attention Network (SAN) to discover product functions from questions. This model leverages unlabeled data as contextual information to perform semi-supervised sequence labeling. We conduct experiments to show that the extracted function have both high coverage and accuracy, compared with a wide spectrum of baselines.

Hu Xu, Sihong Xie, Lei Shu et al.

Product compatibility and their functionality are of utmost importance to customers when they purchase products, and to sellers and manufacturers when they sell products. Due to the huge number of products available online, it is infeasible to enumerate and test the compatibility and functionality of every product. In this paper, we address two closely related problems: product compatibility analysis and function satisfiability analysis, where the second problem is a generalization of the first problem (e.g., whether a product works with another product can be considered as a special function). We first identify a novel question and answering corpus that is up-to-date regarding product compatibility and functionality information. To allow automatic discovery product compatibility and functionality, we then propose a deep learning model called Dual Attention Network (DAN). Given a QA pair for a to-be-purchased product, DAN learns to 1) discover complementary products (or functions), and 2) accurately predict the actual compatibility (or satisfiability) of the discovered products (or functions). The challenges addressed by the model include the briefness of QAs, linguistic patterns indicating compatibility, and the appropriate fusion of questions and answers. We conduct experiments to quantitatively and qualitatively show that the identified products and functions have both high coverage and accuracy, compared with a wide spectrum of baselines.

Lei Shu, Hu Xu, Bing Liu

Traditional supervised learning makes the closed-world assumption that the classes appeared in the test data must have appeared in training. This also applies to text learning or text classification. As learning is used increasingly in dynamic open environments where some new/test documents may not belong to any of the training classes, identifying these novel documents during classification presents an important problem. This problem is called open-world classification or open classification. This paper proposes a novel deep learning based approach. It outperforms existing state-of-the-art techniques dramatically.

Hu Xu, Lei Shu, Philip S. Yu

Extracting opinion targets is an important task in sentiment analysis on product reviews and complementary entities (products) are one important type of opinion targets that may work together with the reviewed product. In this paper, we address the problem of Complementary Entity Recognition (CER) as a supervised sequence labeling with the capability of expanding domain knowledge as key-value pairs from unlabeled reviews, by automatically learning and enhancing knowledge-based features. We use Conditional Random Field (CRF) as the base learner and augment CRF with knowledge-based features (called the Knowledge-based CRF or KCRF for short). We conduct experiments to show that KCRF effectively improves the performance of supervised CER task.

Lei Shu, Hu Xu, Bing Liu

This paper makes a focused contribution to supervised aspect extraction. It shows that if the system has performed aspect extraction from many past domains and retained their results as knowledge, Conditional Random Fields (CRF) can leverage this knowledge in a lifelong learning manner to extract in a new domain markedly better than the traditional CRF without using this prior knowledge. The key innovation is that even after CRF training, the model can still improve its extraction with experiences in its applications.

Lei Shu, Bing Liu, Hu Xu et al.

One of the key tasks of sentiment analysis of product reviews is to extract product aspects or features that users have expressed opinions on. In this work, we focus on using supervised sequence labeling as the base approach to performing the task. Although several extraction methods using sequence labeling methods such as Conditional Random Fields (CRF) and Hidden Markov Models (HMM) have been proposed, we show that this supervised approach can be significantly improved by exploiting the idea of concept sharing across multiple domains. For example, "screen" is an aspect in iPhone, but not only iPhone has a screen, many electronic devices have screens too. When "screen" appears in a review of a new domain (or product), it is likely to be an aspect too. Knowing this information enables us to do much better extraction in the new domain. This paper proposes a novel extraction method exploiting this idea in the context of supervised sequence labeling. Experimental results show that it produces markedly better results than without using the past information.

Hu Xu, Lei Shu, Jingyuan Zhang et al.

Product Community Question Answering (PCQA) provides useful information about products and their features (aspects) that may not be well addressed by product descriptions and reviews. We observe that a product's compatibility issues with other products are frequently discussed in PCQA and such issues are more frequently addressed in accessories, i.e., via a yes/no question "Does this mouse work with windows 10?". In this paper, we address the problem of extracting compatible and incompatible products from yes/no questions in PCQA. This problem can naturally have a two-stage framework: first, we perform Complementary Entity (product) Recognition (CER) on yes/no questions; second, we identify the polarities of yes/no answers to assign the complementary entities a compatibility label (compatible, incompatible or unknown). We leverage an existing unsupervised method for the first stage and a 3-class classifier by combining a distant PU-learning method (learning from positive and unlabeled examples) together with a binary classifier for the second stage. The benefit of using distant PU-learning is that it can help to expand more implicit yes/no answers without using any human annotated data. We conduct experiments on 4 products to show that the proposed method is effective.