Zheng Zhang

Xiang Li, Yiqun Yao, Xin Jiang et al. · tencent-ai, tsinghua

Large language models (LLMs) are considered important approaches towards foundational machine intelligence, achieving remarkable success in Natural Language Processing and multimodal tasks, among others. However, the carbon footprints and financial costs originating from heavy pre-training computation is a non-negligible issue. Progressive training methods, inspired by the neurogenesis process that grows neural structures, have shown potential to accelerate LLM pre-training. However, the algorithms, implementation, and practices for progressively training LLMs beyond 100B parameters remain underexplored. In this paper, we show that our model, namely FLM-101B, trained with our growth strategy under a budget of \$100K, reaches 80\% of the baselines' performances with only 10\% of their floating-point operations. We believe that further studies on progressive training will benefit the community by cutting down the costs and promoting green AI. The checkpoint of FLM-101B is released at https://huggingface.co/CofeAI/FLM-101B.

Cheng Jiayang, Lin Qiu, Tsz Ho Chan et al. · tencent-ai

Analogy-making between narratives is crucial for human reasoning. In this paper, we evaluate the ability to identify and generate analogies by constructing a first-of-its-kind large-scale story-level analogy corpus, \textsc{StoryAnalogy}, which contains 24K story pairs from diverse domains with human annotations on two similarities from the extended Structure-Mapping Theory. We design a set of tests on \textsc{StoryAnalogy}, presenting the first evaluation of story-level analogy identification and generation. Interestingly, we find that the analogy identification tasks are incredibly difficult not only for sentence embedding models but also for the recent large language models (LLMs) such as ChatGPT and LLaMa. ChatGPT, for example, only achieved around 30% accuracy in multiple-choice questions (compared to over 85% accuracy for humans). Furthermore, we observe that the data in \textsc{StoryAnalogy} can improve the quality of analogy generation in LLMs, where a fine-tuned FlanT5-xxl model achieves comparable performance to zero-shot ChatGPT.

Tianhang Zhang, Lin Qiu, Qipeng Guo et al.

Large Language Models (LLMs) have gained significant popularity for their impressive performance across diverse fields. However, LLMs are prone to hallucinate untruthful or nonsensical outputs that fail to meet user expectations in many real-world applications. Existing works for detecting hallucinations in LLMs either rely on external knowledge for reference retrieval or require sampling multiple responses from the LLM for consistency verification, making these methods costly and inefficient. In this paper, we propose a novel reference-free, uncertainty-based method for detecting hallucinations in LLMs. Our approach imitates human focus in factuality checking from three aspects: 1) focus on the most informative and important keywords in the given text; 2) focus on the unreliable tokens in historical context which may lead to a cascade of hallucinations; and 3) focus on the token properties such as token type and token frequency. Experimental results on relevant datasets demonstrate the effectiveness of our proposed method, which achieves state-of-the-art performance across all the evaluation metrics and eliminates the need for additional information.

Zheng Zhang, Hossein Amiri, Zhenke Liu et al.

Identifying anomalous human spatial trajectory patterns can indicate dynamic changes in mobility behavior with applications in domains like infectious disease monitoring and elderly care. Recent advancements in large language models (LLMs) have demonstrated their ability to reason in a manner akin to humans. This presents significant potential for analyzing temporal patterns in human mobility. In this paper, we conduct empirical studies to assess the capabilities of leading LLMs like GPT-4 and Claude-2 in detecting anomalous behaviors from mobility data, by comparing to specialized methods. Our key findings demonstrate that LLMs can attain reasonable anomaly detection performance even without any specific cues. In addition, providing contextual clues about potential irregularities could further enhances their prediction efficacy. Moreover, LLMs can provide reasonable explanations for their judgments, thereby improving transparency. Our work provides insights on the strengths and limitations of LLMs for human spatial trajectory analysis.

Xiangkun Hu, Dongyu Ru, Lin Qiu et al.

Large Language Models (LLMs) have shown impressive capabilities but also a concerning tendency to hallucinate. This paper presents RefChecker, a framework that introduces claim-triplets to represent claims in LLM responses, aiming to detect fine-grained hallucinations. In RefChecker, an extractor generates claim-triplets from a response, which are then evaluated by a checker against a reference. We delineate three task settings: Zero, Noisy and Accurate Context, to reflect various real-world use cases. We curated a benchmark spanning various NLP tasks and annotated 11k claim-triplets from 2.1k responses by seven LLMs. RefChecker supports both proprietary and open-source models as the extractor and checker. Experiments demonstrate that claim-triplets enable superior hallucination detection, compared to other granularities such as response, sentence and sub-sentence level claims. RefChecker outperforms prior methods by 6.8 to 26.1 points on our benchmark and the checking results of RefChecker are strongly aligned with human judgments. This work is open sourced at https://github.com/amazon-science/RefChecker

Zheng Zhang, Wenjie Ai, Kevin Wells et al.

With the development of Human-AI Collaboration in Classification (HAI-CC), integrating users and AI predictions becomes challenging due to the complex decision-making process. This process has three options: 1) AI autonomously classifies, 2) learning to complement, where AI collaborates with users, and 3) learning to defer, where AI defers to users. Despite their interconnected nature, these options have been studied in isolation rather than as components of a unified system. In this paper, we address this weakness with the novel HAI-CC methodology, called Learning to Complement and to Defer to Multiple Users (LECODU). LECODU not only combines learning to complement and learning to defer strategies, but it also incorporates an estimation of the optimal number of users to engage in the decision process. The training of LECODU maximises classification accuracy and minimises collaboration costs associated with user involvement. Comprehensive evaluations across real-world and synthesized datasets demonstrate LECODU's superior performance compared to state-of-the-art HAI-CC methods. Remarkably, even when relying on unreliable users with high rates of label noise, LECODU exhibits significant improvement over both human decision-makers alone and AI alone.

Zheng Zhang, Cuong Nguyen, Kevin Wells et al.

Real-world image classification tasks tend to be complex, where expert labellers are sometimes unsure about the classes present in the images, leading to the issue of learning with noisy labels (LNL). The ill-posedness of the LNL task requires the adoption of strong assumptions or the use of multiple noisy labels per training image, resulting in accurate models that work well in isolation but fail to optimise human-AI collaborative classification (HAI-CC). Unlike such LNL methods, HAI-CC aims to leverage the synergies between human expertise and AI capabilities but requires clean training labels, limiting its real-world applicability. This paper addresses this gap by introducing the innovative Learning to Complement with Multiple Humans (LECOMH) approach. LECOMH is designed to learn from noisy labels without depending on clean labels, simultaneously maximising collaborative accuracy while minimising the cost of human collaboration, measured by the number of human expert annotations required per image. Additionally, new benchmarks featuring multiple noisy labels for both training and testing are proposed to evaluate HAI-CC methods. Through quantitative comparisons on these benchmarks, LECOMH consistently outperforms competitive HAI-CC approaches, human labellers, multi-rater learning, and noisy-label learning methods across various datasets, offering a promising solution for addressing real-world image classification challenges.

Cheng Jiayang, Lin Qiu, Chunkit Chan et al.

Narrative reasoning relies on the understanding of eventualities in story contexts, which requires a wealth of background world knowledge. To help machines leverage such knowledge, existing solutions can be categorized into two groups. Some focus on implicitly modeling eventuality knowledge by pretraining language models (LMs) with eventuality-aware objectives. However, this approach breaks down knowledge structures and lacks interpretability. Others explicitly collect world knowledge of eventualities into structured eventuality-centric knowledge graphs (KGs). However, existing research on leveraging these knowledge sources for free-texts is limited. In this work, we propose an initial comprehensive framework called EventGround, which aims to tackle the problem of grounding free-texts to eventuality-centric KGs for contextualized narrative reasoning. We identify two critical problems in this direction: the event representation and sparsity problems. We provide simple yet effective parsing and partial information extraction methods to tackle these problems. Experimental results demonstrate that our approach consistently outperforms baseline models when combined with graph neural network (GNN) or large language model (LLM) based graph reasoning models. Our framework, incorporating grounded knowledge, achieves state-of-the-art performance while providing interpretable evidence.

Zheng Zhang, Fan Yang, Ziyan Jiang et al. · amazon-science

Recent advances in large language models (LLMs) have enhanced their ability to process long input contexts. This development is particularly crucial for tasks that involve retrieving knowledge from an external datastore, which can result in long inputs. However, recent studies show a positional bias in LLMs, demonstrating varying performance depending on the location of useful information within the input sequence. In this study, we conduct extensive experiments to investigate the root causes of positional bias. Our findings indicate that the primary contributor to LLM positional bias stems from the inherent positional preferences of different models. We demonstrate that merely employing prompt-based solutions is inadequate for overcoming the positional preferences. To address this positional bias issue of a pre-trained LLM, we developed a Position-Aware Parameter Efficient Fine-Tuning (PAPEFT) approach which is composed of a data augmentation technique and a parameter efficient adapter, enhancing a uniform attention distribution across the input context. Our experiments demonstrate that the proposed approach effectively reduces positional bias, improving LLMs' effectiveness in handling long context sequences for various tasks that require externally retrieved knowledge.

Longwei Wang, Xueqian Li, Zheng Zhang

The resilience of convolutional neural networks against input variations and adversarial attacks remains a significant challenge in image recognition tasks. Motivated by the need for more robust and reliable image recognition systems, we propose the Dense Cross-Connected Ensemble Convolutional Neural Network (DCC-ECNN). This novel architecture integrates the dense connectivity principle of DenseNet with the ensemble learning strategy, incorporating intermediate cross-connections between different DenseNet paths to facilitate extensive feature sharing and integration. The DCC-ECNN architecture leverages DenseNet's efficient parameter usage and depth while benefiting from the robustness of ensemble learning, ensuring a richer and more resilient feature representation.

Sajjad Ghiasvand, Yifan Yang, Zhiyu Xue et al.

Parameter-efficient fine-tuning (PEFT) methods typically assume that Large Language Models (LLMs) are trained on data from a single device or client. However, real-world scenarios often require fine-tuning these models on private data distributed across multiple devices. Federated Learning (FL) offers an appealing solution by preserving user privacy, as sensitive data remains on local devices during training. Nonetheless, integrating PEFT methods into FL introduces two main challenges: communication overhead and data heterogeneity. In this paper, we introduce FedTT and FedTT+, methods for adapting LLMs by integrating tensorized adapters into client-side models' encoder/decoder blocks. FedTT is versatile and can be applied to both cross-silo FL and large-scale cross-device FL. FedTT+, an extension of FedTT tailored for cross-silo FL, enhances robustness against data heterogeneity by adaptively freezing portions of tensor factors, further reducing the number of trainable parameters. Experiments on BERT and LLaMA models demonstrate that our proposed methods successfully address data heterogeneity challenges and perform on par or even better than existing federated PEFT approaches while achieving up to 10$\times$ reduction in communication cost.

Zheng Zhang, Cuong Nguyen, Kevin Wells et al.

Human-AI cooperative classification (HAI-CC) approaches aim to develop hybrid intelligent systems that enhance decision-making in various high-stakes real-world scenarios by leveraging both human expertise and AI capabilities. Current HAI-CC methods primarily focus on learning-to-defer (L2D), where decisions are deferred to human experts, and learning-to-complement (L2C), where AI and human experts make predictions cooperatively. However, a notable research gap remains in effectively exploring both L2D and L2C under diverse expert knowledge to improve decision-making, particularly when constrained by the cooperation cost required to achieve a target probability for AI-only selection (i.e., coverage). In this paper, we address this research gap by proposing the Coverage-constrained Learning to Defer and Complement with Specific Experts (CL2DC) method. CL2DC makes final decisions through either AI prediction alone or by deferring to or complementing a specific expert, depending on the input data. Furthermore, we propose a coverage-constrained optimisation to control the cooperation cost, ensuring it approximates a target probability for AI-only selection. This approach enables an effective assessment of system performance within a specified budget. Also, CL2DC is designed to address scenarios where training sets contain multiple noisy-label annotations without any clean-label references. Comprehensive evaluations on both synthetic and real-world datasets demonstrate that CL2DC achieves superior performance compared to state-of-the-art HAI-CC methods.

Guangyu Shen, Zhihua Li, Xiang Xu et al. · amazon-science

Existing deepfake detection techniques struggle to keep-up with the ever-evolving novel, unseen forgeries methods. This limitation stems from their reliance on statistical artifacts learned during training, which are often tied to specific generation processes that may not be representative of samples from new, unseen deepfake generation methods encountered at test time. We propose that incorporating language guidance can improve deepfake detection generalization by integrating human-like commonsense reasoning -- such as recognizing logical inconsistencies and perceptual anomalies -- alongside statistical cues. To achieve this, we train an expert deepfake vision encoder by combining discriminative classification with image-text contrastive learning, where the text is generated by generalist MLLMs using few-shot prompting. This allows the encoder to extract both language-describable, commonsense deepfake artifacts and statistical forgery artifacts from pixel-level distributions. To further enhance robustness, we integrate data uncertainty learning into vision-language contrastive learning, mitigating noise in image-text supervision. Our expert vision encoder seamlessly interfaces with an LLM, further enabling more generalized and interpretable deepfake detection while also boosting accuracy. The resulting framework, AuthGuard, achieves state-of-the-art deepfake detection accuracy in both in-distribution and out-of-distribution settings, achieving AUC gains of 6.15% on the DFDC dataset and 16.68% on the DF40 dataset. Additionally, AuthGuard significantly enhances deepfake reasoning, improving performance by 24.69% on the DDVQA dataset.