As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user's preferred answer ("sycophancy") and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors.

As AI systems become more capable, we would like to enlist their help to supervise other AIs. We experiment with methods for training a harmless AI assistant through self-improvement, without any human labels identifying harmful outputs. The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'. The process involves both a supervised learning and a reinforcement learning phase. In the supervised phase we sample from an initial model, then generate self-critiques and revisions, and then finetune the original model on revised responses. In the RL phase, we sample from the finetuned model, use a model to evaluate which of the two samples is better, and then train a preference model from this dataset of AI preferences. We then train with RL using the preference model as the reward signal, i.e. we use 'RL from AI Feedback' (RLAIF). As a result we are able to train a harmless but non-evasive AI assistant that engages with harmful queries by explaining its objections to them. Both the SL and RL methods can leverage chain-of-thought style reasoning to improve the human-judged performance and transparency of AI decision making. These methods make it possible to control AI behavior more precisely and with far fewer human labels.

Surgery is the only viable treatment for cataract patients with visual acuity (VA) impairment. Clinically, to assess the necessity of cataract surgery, accurately predicting postoperative VA before surgery by analyzing multi-view optical coherence tomography (OCT) images is crucially needed. Unfortunately, due to complicated fundus conditions, determining postoperative VA remains difficult for medical experts. Deep learning methods for this problem were developed in recent years. Although effective, these methods still face several issues, such as not efficiently exploring potential relations between multi-view OCT images, neglecting the key role of clinical prior knowledge (e.g., preoperative VA value), and using only regression-based metrics which are lacking reference. In this paper, we propose a novel Cross-token Transformer Network (CTT-Net) for postoperative VA prediction by analyzing both the multi-view OCT images and preoperative VA. To effectively fuse multi-view features of OCT images, we develop cross-token attention that could restrict redundant/unnecessary attention flow. Further, we utilize the preoperative VA value to provide more information for postoperative VA prediction and facilitate fusion between views. Moreover, we design an auxiliary classification loss to improve model performance and assess VA recovery more sufficiently, avoiding the limitation by only using the regression metrics. To evaluate CTT-Net, we build a multi-view OCT image dataset collected from our collaborative hospital. A set of extensive experiments validate the effectiveness of our model compared to existing methods in various metrics. Code is available at: https://github.com/wjh892521292/Cataract OCT.

Recent development of deep neural networks (DNNs) for tabular learning has largely benefited from the capability of DNNs for automatic feature interaction. However, the heterogeneity nature of tabular features makes such features relatively independent, and developing effective methods to promote tabular feature interaction still remains an open problem. In this paper, we propose a novel Graph Estimator, which automatically estimates the relations among tabular features and builds graphs by assigning edges between related features. Such relation graphs organize independent tabular features into a kind of graph data such that interaction of nodes (tabular features) can be conducted in an orderly fashion. Based on our proposed Graph Estimator, we present a bespoke Transformer network tailored for tabular learning, called T2G-Former, which processes tabular data by performing tabular feature interaction guided by the relation graphs. A specific Cross-level Readout collects salient features predicted by the layers in T2G-Former across different levels, and attains global semantics for final prediction. Comprehensive experiments show that our T2G-Former achieves superior performance among DNNs and is competitive with non-deep Gradient Boosted Decision Tree models.

“感应头”是注意力头，它实现了一种简单的算法来完成令牌序列，例如[a] [b] ... [a] - > [b]。在这项工作中，我们提供了一个假设的初步和间接证据，即诱导头可能构成大型大型变压器模型中所有“文本学习”中大多数的机制（即减少在增加代币指数时损失的损失）。我们发现，诱导头在与秘密学习能力突然急剧上的急剧上升的位置完全相同，这是训练损失的颠簸。我们提出了六种互补的证据，认为诱导头可能是任何大小的变压器模型中一般性内部学习的机理来源。对于仅关注的小型模型，我们提供了有力的因果证据。对于具有MLP的较大模型，我们提供相关证据。

近年来，自动对色素，非色素和脱发的非胸膜皮肤病变的分类引起了很多关注。但是，皮肤纹理，病变形状，脱位对比度，照明条件等的成像变化。阻碍了鲁棒的特征提取，从而影响分类精度。在本文中，我们提出了一个新的深神经网络，该网络利用输入数据进行鲁棒特征提取。具体而言，我们分析了卷积网络的行为（视野），以找到深度监督的位置，以改善特征提取。为了实现这一目标，首先，我们执行激活映射以生成对象掩码，突出显示对分类输出生成最重要的输入区域。然后，选择层的有效接收场的网络层与对象掩模中的近似对象形状相匹配，以作为我们进行深度监督的焦点。利用三个黑色素瘤检测数据集和两个白癜风检测数据集上的不同类型的卷积特征提取器和分类器，我们验证了新方法的有效性。

心电图（ECG）是心脏病的广泛使用的非侵入性诊断工具。许多研究设计了ECG分析模型（例如分类器）来协助诊断。作为一项上游任务，研究建立了生成模型来综合ECG数据，这对提供培训样本，隐私保护和减少注释是有益的。但是，以前的ECG生成方法通常既不合成多视图数据，也不涉及心脏病状况。在本文中，我们提出了一种新型的，用于多视图ECG合成的新型疾病的生成对抗网络，称为ME-GAN，该网络获得了以心脏病为条件的全磁心电图表示，并将其投射到多个标准视图上，以产生ECG信号。由于心脏病的心电图表现通常位于特定波形中，因此我们提出了一种新的“混合标准化”，以精确地注入合适的位置。此外，我们提出了一个视图歧视者，将无序的心电图视图恢复为预定的顺序，监督发电机以获取代表正确视图特征的ECG。此外，提出了一个新的度量RFID，以评估合成的ECG信号的质量。全面的实验验证了我们的ME-GAN在具有可信赖的病态表现的多视图ECG信号合成上表现良好。

我们研究语言模型是否可以评估自己主张的有效性，并预测他们能够正确回答的问题。我们首先表明，当以正确的格式提供时，较大的模型在多样化的多项选择和True/False问题上进行了很好的校准。因此，我们可以通过要求模型首先提出答案，然后评估其答案正确的概率“ p（true）”来对开放式采样任务进行自我评估。我们发现在各种任务中，P（true）的表现，校准和缩放令人鼓舞。当我们允许模型考虑自己的许多样本之前，在预测一种特定可能性的有效性之前，自我评估的性能进一步改善。接下来，我们研究是否可以培训模型来预测“ P（ik）”，即“我知道”问题的概率，而无需参考任何特定提出的答案。模型在预测P（IK）方面表现良好，并且在跨任务中部分概括，尽管它们在新任务上的P（IK）校准方面遇到了困难。预测的p（IK）概率在存在相关的原始材料的情况下以及对数学单词问题解决方案的提示也适当增加。我们希望这些观察结果为培训更诚实的模型提供了基础，并研究了诚实对模型模仿人类写作以外的其他目标培训的案例的普遍性。

我们旨在定量衡量医学图像分割模型的实际可用性：可以使用/信任模型的预测在多大程度上，多久和在哪些样品上进行样本。我们首先提出了一个度量，正确的信心等级相关性（CCRC），以捕获预测的置信度估计如何与其正确性分数相关。具有高价值CCRC的模型意味着其预测信心可靠地表明，哪些样本的预测更可能是正确的。由于CCRC没有捕获实际的预测正确性，因此仅仅指示预测模型是否既准确又可靠地用于实践中。因此，我们进一步提出了另一种可用区域估计（URE）的方法，同时量化了预测在一个估计中的置信度评估的正确性和可靠性。 URE提供了有关模型的预测在多大程度上可用的具体信息。此外，可以利用可用区域（UR）的大小来比较模型：具有较大UR的模型可以作为更可用的模型，因此可以将其视为更好的模型。六个数据集的实验验证了所提出的评估方法表现良好，为医学图像分割模型的实际可用性提供了具体和简洁的措施。代码可在https://github.com/yizhezhang2000/ure上提供。

语言模型既展示了定量的改进，又展示了新的定性功能，随着规模的增加。尽管它们具有潜在的变革性影响，但这些新能力的特征却很差。为了为未来的研究提供信息，为破坏性的新模型能力做准备，并改善社会有害的效果，至关重要的是，我们必须了解目前和近乎未来的能力和语言模型的局限性。为了应对这一挑战，我们介绍了超越模仿游戏基准（Big Bench）。 Big Bench目前由204个任务组成，由132家机构的442位作者贡献。任务主题是多样的，从语言学，儿童发展，数学，常识性推理，生物学，物理学，社会偏见，软件开发等等。 Big-Bench专注于被认为超出当前语言模型的功能的任务。我们评估了OpenAI的GPT型号，Google内部密集变压器体系结构和大型基础上的开关稀疏变压器的行为，跨越了数百万到数十亿个参数。此外，一个人类专家评估者团队执行了所有任务，以提供强大的基准。研究结果包括：模型性能和校准都随规模改善，但绝对的术语（以及与评估者的性能相比）；在模型类中的性能非常相似，尽管带有稀疏性。逐渐和预测的任务通常涉及大量知识或记忆成分，而在临界规模上表现出“突破性”行为的任务通常涉及多个步骤或组成部分或脆性指标；社交偏见通常会随着含糊不清的环境而随着规模而增加，但这可以通过提示来改善。