Evaluating large language models (LLMs) is fundamental, particularly in the context of practical applications. Conventional evaluation methods, typically designed primarily for LLM development, yield numerical scores that ignore the user experience. Therefore, our study shifts the focus from model-centered to human-centered evaluation in the context of AI-powered writing assistance applications. Our proposed metric, termed ``Revision Distance,'' utilizes LLMs to suggest revision edits that mimic the human writing process. It is determined by counting the revision edits generated by LLMs. Benefiting from the generated revision edit details, our metric can provide a self-explained text evaluation result in a human-understandable manner beyond the context-independent score. Our results show that for the easy-writing task, ``Revision Distance'' is consistent with established metrics (ROUGE, Bert-score, and GPT-score), but offers more insightful, detailed feedback and better distinguishes between texts. Moreover, in the context of challenging academic writing tasks, our metric still delivers reliable evaluations where other metrics tend to struggle. Furthermore, our metric also holds significant potential for scenarios lacking reference texts.
Efficient fine-tuning is vital for adapting large language models (LLMs) to downstream tasks. However, it requires non-trivial efforts to implement these methods on different models. We present LlamaFactory, a unified framework that integrates a suite of cutting-edge efficient training methods. It allows users to flexibly customize the fine-tuning of 100+ LLMs without the need for coding through the built-in web UI LlamaBoard. We empirically validate the efficiency and effectiveness of our framework on language modeling and text generation tasks. It has been released at https://github.com/hiyouga/LLaMA-Factory and already received over 13,000 stars and 1,600 forks.
Large Vision-Language Models (LVLMs) have shown significant progress in well responding to visual-instructions from users. However, these instructions, encompassing images and text, are susceptible to both intentional and inadvertent attacks. Despite the critical importance of LVLMs' robustness against such threats, current research in this area remains limited. To bridge this gap, we introduce AVIBench, a framework designed to analyze the robustness of LVLMs when facing various adversarial visual-instructions (AVIs), including four types of image-based AVIs, ten types of text-based AVIs, and nine types of content bias AVIs (such as gender, violence, cultural, and racial biases, among others). We generate 260K AVIs encompassing five categories of multimodal capabilities (nine tasks) and content bias. We then conduct a comprehensive evaluation involving 14 open-source LVLMs to assess their performance. AVIBench also serves as a convenient tool for practitioners to evaluate the robustness of LVLMs against AVIs. Our findings and extensive experimental results shed light on the vulnerabilities of LVLMs, and highlight that inherent biases exist even in advanced closed-source LVLMs like GeminiProVision and GPT-4V. This underscores the importance of enhancing the robustness, security, and fairness of LVLMs. The source code and benchmark will be made publicly available.
Deciphering visual content from functional Magnetic Resonance Imaging (fMRI) helps illuminate the human vision system. However, the scarcity of fMRI data and noise hamper brain decoding model performance. Previous approaches primarily employ subject-specific models, sensitive to training sample size. In this paper, we explore a straightforward but overlooked solution to address data scarcity. We propose shallow subject-specific adapters to map cross-subject fMRI data into unified representations. Subsequently, a shared deeper decoding model decodes cross-subject features into the target feature space. During training, we leverage both visual and textual supervision for multi-modal brain decoding. Our model integrates a high-level perception decoding pipeline and a pixel-wise reconstruction pipeline guided by high-level perceptions, simulating bottom-up and top-down processes in neuroscience. Empirical experiments demonstrate robust neural representation learning across subjects for both pipelines. Moreover, merging high-level and low-level information improves both low-level and high-level reconstruction metrics. Additionally, we successfully transfer learned general knowledge to new subjects by training new adapters with limited training data. Compared to previous state-of-the-art methods, notably pre-training-based methods (Mind-Vis and fMRI-PTE), our approach achieves comparable or superior results across diverse tasks, showing promise as an alternative method for cross-subject fMRI data pre-training. Our code and pre-trained weights will be publicly released at https://github.com/YulongBonjour/See_Through_Their_Minds.
The distinction between humans and animals lies in the unique ability of humans to use and create tools. Tools empower humans to overcome physiological limitations, fostering the creation of magnificent civilizations. Similarly, enabling foundational models like Large Language Models (LLMs) with the capacity to learn external tool usage may serve as a pivotal step toward realizing artificial general intelligence. Previous studies in this field have predominantly pursued two distinct approaches to augment the tool invocation capabilities of LLMs. The first approach emphasizes the construction of relevant datasets for model fine-tuning. The second approach, in contrast, aims to fully exploit the inherent reasoning abilities of LLMs through in-context learning strategies. In this work, we introduce a novel tool invocation pipeline designed to control massive real-world APIs. This pipeline mirrors the human task-solving process, addressing complicated real-life user queries. At each step, we guide LLMs to summarize the achieved results and determine the next course of action. We term this pipeline `from Summary to action', Sum2Act for short. Empirical evaluations of our Sum2Act pipeline on the ToolBench benchmark show significant performance improvements, outperforming established methods like ReAct and DFSDT. This highlights Sum2Act's effectiveness in enhancing LLMs for complex real-world tasks.
Fine-tuning pre-trained language models (PLMs), e.g., SciBERT, generally requires large numbers of annotated data to achieve state-of-the-art performance on a range of NLP tasks in the scientific domain. However, obtaining the fine-tune data for scientific NLP task is still challenging and expensive. Inspired by recent advancement in prompt learning, in this paper, we propose the Mix Prompt Tuning (MPT), which is a semi-supervised method to alleviate the dependence on annotated data and improve the performance of multi-granularity academic function recognition tasks with a small number of labeled examples. Specifically, the proposed method provides multi-perspective representations by combining manual prompt templates with automatically learned continuous prompt templates to help the given academic function recognition task take full advantage of knowledge in PLMs. Based on these prompt templates and the fine-tuned PLM, a large number of pseudo labels are assigned to the unlabeled examples. Finally, we fine-tune the PLM using the pseudo training set. We evaluate our method on three academic function recognition tasks of different granularity including the citation function, the abstract sentence function, and the keyword function, with datasets from computer science domain and biomedical domain. Extensive experiments demonstrate the effectiveness of our method and statistically significant improvements against strong baselines. In particular, it achieves an average increase of 5% in Macro-F1 score compared with fine-tuning, and 6% in Macro-F1 score compared with other semi-supervised method under low-resource settings. In addition, MPT is a general method that can be easily applied to other low-resource scientific classification tasks.
Reconstructing perceived natural images or decoding their categories from fMRI signals are challenging tasks with great scientific significance. Due to the lack of paired samples, most existing methods fail to generate semantically recognizable reconstruction and are difficult to generalize to novel classes. In this work, we propose, for the first time, a task-agnostic brain decoding model by unifying the visual stimulus classification and reconstruction tasks in a semantic space. We denote it as BrainCLIP, which leverages CLIP's cross-modal generalization ability to bridge the modality gap between brain activities, images, and texts. Specifically, BrainCLIP is a VAE-based architecture that transforms fMRI patterns into the CLIP embedding space by combining visual and textual supervision. Note that previous works rarely use multi-modal supervision for visual stimulus decoding. Our experiments demonstrate that textual supervision can significantly boost the performance of decoding models compared to the condition where only image supervision exists. BrainCLIP can be applied to multiple scenarios like fMRI-to-image generation, fMRI-image-matching, and fMRI-text-matching. Compared with BraVL, a recently proposed multi-modal method for fMRI-based brain decoding, BrainCLIP achieves significantly better performance on the novel class classification task. BrainCLIP also establishes a new state-of-the-art for fMRI-based natural image reconstruction in terms of high-level image features.
BACKGROUND: Recent neural language models have taken a significant step forward in producing remarkably controllable, fluent, and grammatical text. Although some recent works have found that AI-generated text is not distinguishable from human-authored writing for crowd-sourcing workers, there still exist errors in AI-generated text which are even subtler and harder to spot. METHOD: In this paper, we investigate the gap between scientific content generated by AI and written by humans. Specifically, we first adopt several publicly available tools or models to investigate the performance for detecting GPT-generated scientific text. Then we utilize features from writing style to analyze the similarities and differences between the two types of content. Furthermore, more complex and deep perspectives, such as consistency, coherence, language redundancy, and factual errors, are also taken into consideration for in-depth analysis. RESULT: The results suggest that while AI has the potential to generate scientific content that is as accurate as human-written content, there is still a gap in terms of depth and overall quality. AI-generated scientific content is more likely to contain errors in language redundancy and factual issues. CONCLUSION: We find that there exists a ``writing style'' gap between AI-generated scientific text and human-written scientific text. Moreover, based on the analysis result, we summarize a series of model-agnostic or distribution-agnostic features, which could be utilized to unknown or novel domain distribution and different generation methods. Future research should focus on not only improving the capabilities of AI models to produce high-quality content but also examining and addressing ethical and security concerns related to the generation and the use of AI-generated content.
Brain decoding is a hot spot in cognitive science, which focuses on reconstructing perceptual images from brain activities. Analyzing the correlations of collected data from human brain activities and representing activity patterns are two problems in brain decoding based on functional magnetic resonance imaging (fMRI) signals. However, existing correlation analysis methods mainly focus on the strength information of voxel, which reveals functional connectivity in the cerebral cortex. They tend to neglect the structural information that implies the intracortical or intrinsic connections; that is, structural connectivity. Hence, the effective connectivity inferred by these methods is relatively unilateral. Therefore, we proposed a correlation network (CorrNet) framework that could be flexibly combined with diverse pattern representation models. In the CorrNet framework, the topological correlation was introduced to reveal structural information. Rich correlations were obtained, which contributed to specifying the underlying effective connectivity. We also combined the CorrNet framework with a linear support vector machine (SVM) and a dynamic evolving spike neuron network (SNN) for pattern representation separately, thus providing a novel method for decoding cognitive activity patterns. Experimental results verified the reliability and robustness of our CorrNet framework and demonstrated that the new method achieved significant improvement in brain decoding over comparable methods.