了解强化学习（RL）代理的新兴行为可能很困难，因为这种代理通常使用高度复杂的决策程序在复杂的环境中进行训练。这引起了RL中解释性的多种方法，旨在调和可能在主体行为与观察者预期的行为之间产生的差异。最近的方法取决于域知识，这可能并非总是可用的，分析代理商的策略，或者是对基础环境的特定要素的分析，通常被建模为马尔可夫决策过程（MDP）。我们的主要主张是，即使基本的MDP尚不完全了解（例如，尚未准确地了解过渡概率），也没有由代理商维护（即，在使用无模型方法时），但仍可以利用它为自动生成解释。为此，我们建议使用以前在文献中使用的正式MDP抽象和转换来加快寻找最佳策略的搜索，以自动产生解释。由于这种转换通常基于环境的符号表示，因此它们可能代表了预期和实际代理行为之间差距的有意义的解释。我们正式定义了这个问题，建议一类可用于解释新兴行为的转换，并提出了有效搜索解释的方法。我们演示了一组标准基准测试的方法。

我们考虑战略设置，其中几个用户在重复的在线互动中聘用，辅助最小化的代理商代表他们反复发挥“游戏”。我们研究了代理人的重复游戏的动态和平均结果，并将其视为诱导用户之间的元游戏。我们的主要焦点是用户可以在此元游戏中从“操纵”他们自己的代理商中可以受益于他们自己的代理商。我们正式定义了普通游戏的这种“用户代理元荟萃游戏”模型，讨论了自动化代理动态的不同概念下的属性，并分析了2x2游戏中用户的均衡，其中动态收敛到a单均衡。

我们分析了一种方案，其中软件代理作为后悔最小化算法代表他们的用户参与重复拍卖。我们研究了第一个价格和第二次价格拍卖，以及他们的广义版本（例如，作为用于广告拍卖的版本）。利用理论分析和模拟，我们展示了，令人惊讶的是，在二次价格拍卖中，球员的激励措施将他们的真正估值释放到自己的学习代理，而在第一次价格拍卖中，这是所有球员如实的主要战略向他们的代理商报告他们的估值。

This paper presents a machine learning approach to multidimensional item response theory (MIRT), a class of latent factor models that can be used to model and predict student performance from observed assessment data. Inspired by collaborative filtering, we define a general class of models that includes many MIRT models. We discuss the use of penalized joint maximum likelihood (JML) to estimate individual models and cross-validation to select the best performing model. This model evaluation process can be optimized using batching techniques, such that even sparse large-scale data can be analyzed efficiently. We illustrate our approach with simulated and real data, including an example from a massive open online course (MOOC). The high-dimensional model fit to this large and sparse dataset does not lend itself well to traditional methods of factor interpretation. By analogy to recommender-system applications, we propose an alternative "validation" of the factor model, using auxiliary information about the popularity of items consulted during an open-book exam in the course.

For applications that require processing large amounts of text at inference time, Large Language Models (LLMs) are handicapped by their limited context windows, which are typically 2048 tokens. In-context learning, an emergent phenomenon in LLMs in sizes above a certain parameter threshold, constitutes one significant example because it can only leverage training examples that fit into the context window. Existing efforts to address the context window limitation involve training specialized architectures, which tend to be smaller than the sizes in which in-context learning manifests due to the memory footprint of processing long texts. We present Parallel Context Windows (PCW), a method that alleviates the context window restriction for any off-the-shelf LLM without further training. The key to the approach is to carve a long context into chunks (``windows'') that fit within the architecture, restrict the attention mechanism to apply only within each window, and re-use the positional embeddings among the windows. We test the PCW approach on in-context learning with models that range in size between 750 million and 178 billion parameters, and show substantial improvements for tasks with diverse input and output spaces. Our results motivate further investigation of Parallel Context Windows as a method for applying off-the-shelf LLMs in other settings that require long text sequences.

Attribute-controlled text rewriting, also known as text style-transfer, has a crucial role in regulating attributes and biases of textual training data and a machine generated text. In this work we present SimpleStyle, a minimalist yet effective approach for style-transfer composed of two simple ingredients: controlled denoising and output filtering. Despite the simplicity of our approach, which can be succinctly described with a few lines of code, it is competitive with previous state-of-the-art methods both in automatic and in human evaluation. To demonstrate the adaptability and practical value of our system beyond academic data, we apply SimpleStyle to transfer a wide range of text attributes appearing in real-world textual data from social networks. Additionally, we introduce a novel "soft noising" technique that further improves the performance of our system. We also show that teaching a student model to generate the output of SimpleStyle can result in a system that performs style transfer of equivalent quality with only a single greedy-decoded sample. Finally, we suggest our method as a remedy for the fundamental incompatible baseline issue that holds progress in the field. We offer our protocol as a simple yet strong baseline for works that wish to make incremental advancements in the field of attribute controlled text rewriting.

We study the problem of continually training an instruction-following agent through feedback provided by users during collaborative interactions. During interaction, human users instruct an agent using natural language, and provide realtime binary feedback as they observe the agent's instruction execution. We cast learning as a contextual bandit problem, converting the user feedback to immediate reward. We evaluate through multiple rounds of human-agent interactions, demonstrating 15.4% absolute improvement in instruction execution over time. We also show our approach is robust to several design variations, and that the feedback signal is roughly equivalent to the learning signal of supervised demonstration data.

Light is a complex-valued field. The intensity and phase of the field are affected by imaged objects. However, imaging sensors measure only real-valued non-negative intensities. This results in a nonlinear relation between the measurements and the unknown imaged objects. Moreover, the sensor readouts are corrupted by Poissonian-distributed photon noise. In this work, we seek the most probable object (or clear image), given noisy measurements, that is, maximizing the a-posteriori probability of the sought variables. Hence, we generalize annealed Langevin dynamics, tackling fundamental challenges in optical imaging, including phase recovery and Poisson (photon) denoising. We leverage deep neural networks, not for explicit recovery of the imaged object, but as an approximate gradient for a prior term. We show results on empirical data, acquired by a real experiment. We further show results of simulations.

Pretraining has been shown to scale well with compute, data size and data diversity. Multitask learning trains on a mixture of supervised datasets and produces improved performance compared to self-supervised pretraining. Until now, massively multitask learning required simultaneous access to all datasets in the mixture and heavy compute resources that are only available to well-resourced teams. In this paper, we propose ColD Fusion, a method that provides the benefits of multitask learning but leverages distributed computation and requires limited communication and no sharing of data. Consequentially, ColD Fusion can create a synergistic loop, where finetuned models can be recycled to continually improve the pretrained model they are based on. We show that ColD Fusion yields comparable benefits to multitask pretraining by producing a model that (a) attains strong performance on all of the datasets it was multitask trained on and (b) is a better starting point for finetuning on unseen datasets. We find ColD Fusion outperforms RoBERTa and even previous multitask models. Specifically, when training and testing on 35 diverse datasets, ColD Fusion-based model outperforms RoBERTa by 2.45 points in average without any changes to the architecture.

We introduce KiloGram, a resource for studying abstract visual reasoning in humans and machines. Drawing on the history of tangram puzzles as stimuli in cognitive science, we build a richly annotated dataset that, with >1k distinct stimuli, is orders of magnitude larger and more diverse than prior resources. It is both visually and linguistically richer, moving beyond whole shape descriptions to include segmentation maps and part labels. We use this resource to evaluate the abstract visual reasoning capacities of recent multi-modal models. We observe that pre-trained weights demonstrate limited abstract reasoning, which dramatically improves with fine-tuning. We also observe that explicitly describing parts aids abstract reasoning for both humans and models, especially when jointly encoding the linguistic and visual inputs. KiloGram is available at https://lil.nlp.cornell.edu/kilogram .