Ruoyu Zhang

Minhao Zhang, Ruoyu Zhang, Yanzeng Li et al. · pku

Semantic parsing solves knowledge base (KB) question answering (KBQA) by composing a KB query, which generally involves node extraction (NE) and graph composition (GC) to detect and connect related nodes in a query. Despite the strong causal effects between NE and GC, previous works fail to directly model such causalities in their pipeline, hindering the learning of subtask correlations. Also, the sequence-generation process for GC in previous works induces ambiguity and exposure bias, which further harms accuracy. In this work, we formalize semantic parsing into two stages. In the first stage (graph structure generation), we propose a causal-enhanced table-filler to overcome the issues in sequence-modelling and to learn the internal causalities. In the second stage (relation extraction), an efficient beam-search algorithm is presented to scale complex queries on large-scale KBs. Experiments on LC-QuAD 1.0 indicate that our method surpasses previous state-of-the-arts by a large margin (17%) while remaining time and space efficiency. The code and models are available at https://github.com/AOZMH/Crake.

Duo Wu, Dayou Zhang, Miao Zhang et al.

The high-accuracy and resource-intensive deep neural networks (DNNs) have been widely adopted by live video analytics (VA), where camera videos are streamed over the network to resource-rich edge/cloud servers for DNN inference. Common video encoding configurations (e.g., resolution and frame rate) have been identified with significant impacts on striking the balance between bandwidth consumption and inference accuracy and therefore their adaption scheme has been a focus of optimization. However, previous profiling-based solutions suffer from high profiling cost, while existing deep reinforcement learning (DRL) based solutions may achieve poor performance due to the usage of fixed reward function for training the agent, which fails to craft the application goals in various scenarios. In this paper, we propose ILCAS, the first imitation learning (IL) based configuration-adaptive VA streaming system. Unlike DRL-based solutions, ILCAS trains the agent with demonstrations collected from the expert which is designed as an offline optimal policy that solves the configuration adaption problem through dynamic programming. To tackle the challenge of video content dynamics, ILCAS derives motion feature maps based on motion vectors which allow ILCAS to visually ``perceive'' video content changes. Moreover, ILCAS incorporates a cross-camera collaboration scheme to exploit the spatio-temporal correlations of cameras for more proper configuration selection. Extensive experiments confirm the superiority of ILCAS compared with state-of-the-art solutions, with 2-20.9% improvement of mean accuracy and 19.9-85.3% reduction of chunk upload lag.

Ruoyu Zhang, Yanzeng Li, Yongliang Ma et al. · pku

Prevalent supervised learning methods in natural language processing (NLP) are notoriously data-hungry, which demand large amounts of high-quality annotated data. In practice, acquiring such data is a costly endeavor. Recently, the superior few-shot performance of large language models (LLMs) has propelled the development of dataset generation, where the training data are solely synthesized from LLMs. However, such an approach usually suffers from low-quality issues, and requires orders of magnitude more labeled data to achieve satisfactory performance. To fully exploit the potential of LLMs and make use of massive unlabeled data, we propose LLMaAA, which takes LLMs as annotators and puts them into an active learning loop to determine what to annotate efficiently. To learn robustly with pseudo labels, we optimize both the annotation and training processes: (1) we draw k-NN examples from a small demonstration pool as in-context examples, and (2) we adopt the example reweighting technique to assign training samples with learnable weights. Compared with previous approaches, LLMaAA features both efficiency and reliability. We conduct experiments and analysis on two classic NLP tasks, named entity recognition and relation extraction. With LLMaAA, task-specific models trained from LLM-generated labels can outperform the teacher within only hundreds of annotated examples, which is much more cost-effective than other baselines.

Tianhao Wu, Banghua Zhu, Ruoyu Zhang et al.

Large Language Models (LLMs) can acquire extensive world knowledge through pre-training on large corpora. However, due to exposure to low-quality data, LLMs may exhibit harmful behavior without aligning with human values. The dominant approach for steering LLMs towards beneficial behavior involves Reinforcement Learning with Human Feedback (RLHF), with Proximal Policy Optimization (PPO) serving as the default RL optimizer. Despite its effectiveness, PPO has limitations when optimizing rewards trained from comparison-based loss. Primarily, PPO is not invariant to equivalent reward functions containing identical preference information due to the need to calibrate the reward scale. Additionally, PPO's necessity for token-wise updates introduces complexity in both function approximation and algorithm design compared to trajectory-wise optimization. This paper proposes a new framework, reinforcement learning with relative feedback, and a novel trajectory-wise policy gradient algorithm, Pairwise Proximal Policy Optimization (P3O) that operates directly on comparative rewards. We show theoretically that P3O is invariant to equivalent rewards and avoids the complexity of PPO. Empirical evaluations demonstrate that P3O outperforms PPO in the KL-Reward trade-off and can align with human preferences as well as or better than prior methods. In summary, this work introduces a simpler yet effective approach for aligning LLMs to human preferences through relative feedback.

Minhao Zhang, Yongliang Ma, Yanzeng Li et al. · pku

Incorporating multiple knowledge sources is proven to be beneficial for answering complex factoid questions. To utilize multiple knowledge bases (KB), previous works merge all KBs into a single graph via entity alignment and reduce the problem to question-answering (QA) over the fused KB. In reality, various link relations between KBs might be adopted in QA over multi-KBs. In addition to the identity between the alignable entities (i.e. full link), unalignable entities expressing the different aspects or types of an abstract concept may also be treated identical in a question (i.e. partial link). Hence, the KB fusion in prior works fails to represent all types of links, restricting their ability to comprehend multi-KBs for QA. In this work, we formulate the novel Multi-KB-QA task that leverages the full and partial links among multiple KBs to derive correct answers, a benchmark with diversified link and query types is also constructed to efficiently evaluate Multi-KB-QA performance. Finally, we propose a method for Multi-KB-QA that encodes all link relations in the KB embedding to score and rank candidate answers. Experiments show that our method markedly surpasses conventional KB-QA systems in Multi-KB-QA, justifying the necessity of devising this task.

Hu Cui, Renjing Huang, Ruoyu Zhang et al.

Graph convolutional networks (GCNs) have emerged as a powerful tool for skeleton-based action and gesture recognition, thanks to their ability to model spatial and temporal dependencies in skeleton data. However, existing GCN-based methods face critical limitations: (1) they lack effective spatio-temporal topology modeling that captures dynamic variations in skeletal motion, and (2) they struggle to model multiscale structural relationships beyond local joint connectivity. To address these issues, we propose a novel framework called Dynamic Spatial-Temporal Semantic Awareness Graph Convolutional Network (DSTSA-GCN). DSTSA-GCN introduces three key modules: Group Channel-wise Graph Convolution (GC-GC), Group Temporal-wise Graph Convolution (GT-GC), and Multi-Scale Temporal Convolution (MS-TCN). GC-GC and GT-GC operate in parallel to independently model channel-specific and frame-specific correlations, enabling robust topology learning that accounts for temporal variations. Additionally, both modules employ a grouping strategy to adaptively capture multiscale structural relationships. Complementing this, MS-TCN enhances temporal modeling through group-wise temporal convolutions with diverse receptive fields. Extensive experiments demonstrate that DSTSA-GCN significantly improves the topology modeling capabilities of GCNs, achieving state-of-the-art performance on benchmark datasets for gesture and action recognition, including SHREC17 Track, DHG-14\/28, NTU-RGB+D, and NTU-RGB+D-120.