Language models have been shown to perform better with an increase in scale on a wide variety of tasks via the in-context learning paradigm. In this paper, we investigate the hypothesis that the ability of a large language model to in-context learn-perform a task is not uniformly spread across all of its underlying components. Using a 66 billion parameter language model (OPT-66B) across a diverse set of 14 downstream tasks, we find this is indeed the case: $\sim$70% of attention heads and $\sim$20% of feed forward networks can be removed with minimal decline in task performance. We find substantial overlap in the set of attention heads (un)important for in-context learning across tasks and number of in-context examples. We also address our hypothesis through a task-agnostic lens, finding that a small set of attention heads in OPT-66B score highly on their ability to perform primitive induction operations associated with in-context learning, namely, prefix matching and copying. These induction heads overlap with task-specific important heads, suggesting that induction heads are among the heads capable of more sophisticated behaviors associated with in-context learning. Overall, our study provides several insights that indicate large language models may be under-trained to perform in-context learning and opens up questions on how to pre-train language models to more effectively perform in-context learning.

We consider the problem of multi-agent navigation and collision avoidance when observations are limited to the local neighborhood of each agent. We propose InforMARL, a novel architecture for multi-agent reinforcement learning (MARL) which uses local information intelligently to compute paths for all the agents in a decentralized manner. Specifically, InforMARL aggregates information about the local neighborhood of agents for both the actor and the critic using a graph neural network and can be used in conjunction with any standard MARL algorithm. We show that (1) in training, InforMARL has better sample efficiency and performance than baseline approaches, despite using less information, and (2) in testing, it scales well to environments with arbitrary numbers of agents and obstacles.

我们介绍了一个大规模实验，该实验对编码器进行了预处理，其参数计数范围从700m到9.3b不等，随后蒸馏到较小的型号中，范围为17m-170亿参数，其应用到自然语言理解（NLU）组件（NLU）组件（虚拟助手系统。尽管我们使用70％的口语数据训练，但在对书面形式的跨语性自然语言推论（XNLI）语料库进行评估时，我们的教师模型与XLM-R和MT5相当。我们使用系统中的内域数据对教师模型进行了第二阶段的训练，以提高了3.86％的相对分类，而相对7.01％的插槽填充。我们发现，即使是从我们的2阶段教师模型中提取的170亿参数模型，与仅接受公共数据的2.3B参数老师相比，与2.3B参数老师相比，意图分类更好2.88％，并且7.69％的插槽填充错误率更好（第1阶段），强调了。内域数据对训练的重要性。当使用标记的NLU数据进行离线评估时，我们的17m参数阶段2蒸馏模型的表现分别优于XLM-R碱基（85m Params）和Distillbert（42m Params），分别优于4.23％至6.14％。最后，我们介绍了一个完整的虚拟助手实验平台的结果，在该平台中，我们发现使用经过预训练和蒸馏管道训练的模型超过了从8500万参数教师蒸馏的模型，在自动测量全系统用户不满的自动测量中，从8500万参数教师蒸馏出3.74％-4.91％。

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。

利用自然语言任务描述提示输入已成为一种流行的机制，可以从大规模生成语言模型中引出合理准确的输出，几乎没有内心的监督。这也有助于深入了解语言模型如何纯粹捕获广泛的下游任务的语义，这些任务纯粹是在未标记文本的大规模集团上的自我监督的预训练中。这些模型自然也暴露于许多不良内容，如种族主义和性别歧视语言，并且有限地涉及沿着这些尺寸的模型的认识。在本文中，我们定义和全面评估了这种语言模型如何捕获四项任务的语义：诊断，识别，提取和重新展示。我们为这些任务定义了三个广泛的任务描述：语句，问题和完成，每个类内都有许多词汇变体。我们使用这些类和少量解码方法和少量示例的零任务描述来研究提示每项任务的功效。我们的分析表明，语言模型能够在不同偏差尺寸（例如性别和政治附属）上的不同程度上进行广泛变化的程度。我们相信我们的作品是通过量化当前自我监督目标的限制来实现这种社会学挑战性任务的局限性的重要阶段。

隐性知识，例如常识，是人类对话的关键。当前的神经反应生成（RG）模型经过训练以直接产生响应，省略了未阐明的隐式知识。在本文中，我们介绍了说话之前的思维（TBS），这是一种首先将隐式常识知识（思考）外部化的生成方法（思考），并使用这些知识来产生响应（speak）。我们期望外部化隐式知识可以更有效地学习，产生更多信息的响应，并实现了更多可解释的模型。我们分析了不同的选择，以收集知识一致的对话，代表隐式知识以及知识和对话之间的过渡。经验结果表明，TBS模型在大多数自动指标上优于端到端和知识增强的RG基准，并通过人类注释者评估，产生更有信息，具体和常识性遵循的响应。 TBS还产生了有意义的知识，并且与85 \％左右的对话有关。

Quadruped robots are currently used in industrial robotics as mechanical aid to automate several routine tasks. However, presently, the usage of such a robot in a domestic setting is still very much a part of the research. This paper discusses the understanding and virtual simulation of such a robot capable of detecting and understanding human emotions, generating its gait, and responding via sounds and expression on a screen. To this end, we use a combination of reinforcement learning and software engineering concepts to simulate a quadruped robot that can understand emotions, navigate through various terrains and detect sound sources, and respond to emotions using audio-visual feedback. This paper aims to establish the framework of simulating a quadruped robot that is emotionally intelligent and can primarily respond to audio-visual stimuli using motor or audio response. The emotion detection from the speech was not as performant as ERANNs or Zeta Policy learning, still managing an accuracy of 63.5%. The video emotion detection system produced results that are almost at par with the state of the art, with an accuracy of 99.66%. Due to its "on-policy" learning process, the PPO algorithm was extremely rapid to learn, allowing the simulated dog to demonstrate a remarkably seamless gait across the different cadences and variations. This enabled the quadruped robot to respond to generated stimuli, allowing us to conclude that it functions as predicted and satisfies the aim of this work.

Text-to-text generation models have increasingly become the go-to solution for a wide variety of sequence labeling tasks (e.g., entity extraction and dialog slot filling). While most research has focused on the labeling accuracy, a key aspect -- of vital practical importance -- has slipped through the cracks: understanding model confidence. More specifically, we lack a principled understanding of how to reliably gauge the confidence of a model in its predictions for each labeled span. This paper aims to provide some empirical insights on estimating model confidence for generative sequence labeling. Most notably, we find that simply using the decoder's output probabilities is not the best in realizing well-calibrated confidence estimates. As verified over six public datasets of different tasks, we show that our proposed approach -- which leverages statistics from top-$k$ predictions by a beam search -- significantly reduces calibration errors of the predictions of a generative sequence labeling model.

We consider the task of text generation in language models with constraints specified in natural language. To this end, we first create a challenging benchmark Cognac that provides as input to the model a topic with example text, along with a constraint on text to be avoided. Unlike prior work, our benchmark contains knowledge-intensive constraints sourced from databases like Wordnet and Wikidata, which allows for straightforward evaluation while striking a balance between broad attribute-level and narrow lexical-level controls. We find that even state-of-the-art language models like GPT-3 fail often on this task, and propose a solution to leverage a language model's own internal knowledge to guide generation. Our method, called CognacGen, first queries the language model to generate guidance terms for a specified topic or constraint, and uses the guidance to modify the model's token generation probabilities. We propose three forms of guidance (binary verifier, top-k tokens, textual example), and employ prefix-tuning approaches to distill the guidance to tackle diverse natural language constraints. Through extensive empirical evaluations, we demonstrate that CognacGen can successfully generalize to unseen instructions and outperform competitive baselines in generating constraint conforming text.

The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps.