Two core challenges of alignment are 1) scalable oversight and 2) accounting for the dynamic nature of human values. While solutions like recursive reward modeling address 1), they do not simultaneously account for 2). We sketch a roadmap for a novel algorithmic framework that trains a superhuman reasoning model to decompose complex tasks into subtasks that are still amenable to human-level guidance. Our approach relies on what we call the part-to-complete generalization hypothesis, which states that the alignment of subtask solutions generalizes to the alignment of complete solutions. We advocate for the need to measure this generalization and propose ways to improve it in the future.

Combinatorial optimization problems are prevalent across a wide variety of domains. These problems are often nuanced, their optimal solutions might not be efficiently obtainable, and they may require lots of time and compute resources to solve (they are NP-hard). It follows that the best course of action for solving these problems is to use general optimization algorithm paradigms to quickly and easily develop algorithms that are customized to these problems and can produce good solutions in a reasonable amount of time. In this paper, we present optimizn, a Python library for developing customized optimization algorithms under general optimization algorithm paradigms (simulated annealing, branch and bound). Additionally, optimizn offers continuous training, with which users can run their algorithms on a regular cadence, retain the salient aspects of previous runs, and use them in subsequent runs to potentially produce solutions that get closer and closer to optimality. An earlier version of this paper was peer reviewed and published internally at Microsoft.

This paper addresses the Restricted Longest Common Subsequence (RLCS) problem, an extension of the well-known Longest Common Subsequence (LCS) problem. This problem has significant applications in bioinformatics, particularly for identifying similarities and discovering mutual patterns and important motifs among DNA, RNA, and protein sequences. Building on recent advancements in solving this problem through a general search framework, this paper introduces two novel heuristic approaches designed to enhance the search process by steering it towards promising regions in the search space. The first heuristic employs a probabilistic model to evaluate partial solutions during the search process. The second heuristic is based on a neural network model trained offline using a genetic algorithm. A key aspect of this approach is extracting problem-specific features of partial solutions and the complete problem instance. An effective hybrid method, referred to as the learning beam search, is developed by combining the trained neural network model with a beam search framework. An important contribution of this paper is found in the generation of real-world instances where scientific abstracts serve as input strings, and a set of frequently occurring academic words from the literature are used as restricted patterns. Comprehensive experimental evaluations demonstrate the effectiveness of the proposed approaches in solving the RLCS problem. Finally, an empirical explainability analysis is applied to the obtained results. In this way, key feature combinations and their respective contributions to the success or failure of the algorithms across different problem types are identified.

Humans can perform complex tasks with long-term objectives by planning, reasoning, and forecasting outcomes of actions. For embodied agents to achieve similar capabilities, they must gain knowledge of the environment transferable to novel scenarios with a limited budget of additional trial and error. Learning-based approaches, such as deep RL, can discover and take advantage of inherent regularities and characteristics of the application domain from data, and continuously improve their performances, however at a cost of large amounts of training data. This thesis explores the development of data-driven techniques for spatial reasoning and planning tasks, focusing on enhancing learning efficiency, interpretability, and transferability across novel scenarios. Four key contributions are made. 1) CALVIN, a differential planner that learns interpretable models of the world for long-term planning. It successfully navigated partially observable 3D environments, such as mazes and indoor rooms, by learning the rewards and state transitions from expert demonstrations. 2) SOAP, an RL algorithm that discovers options unsupervised for long-horizon tasks. Options segment a task into subtasks and enable consistent execution of the subtask. SOAP showed robust performances on history-conditional corridor tasks as well as classical benchmarks such as Atari. 3) LangProp, a code optimisation framework using LLMs to solve embodied agent problems that require reasoning by treating code as learnable policies. The framework successfully generated interpretable code with comparable or superior performance to human-written experts in the CARLA autonomous driving benchmark. 4) Voggite, an embodied agent with a vision-to-action transformer backend that solves complex tasks in Minecraft. It achieved third place in the MineRL BASALT Competition by identifying action triggers to segment tasks into multiple stages.

Integrating AI-driven tools in higher education is an emerging area with transformative potential. This paper introduces Iris, a chat-based virtual tutor integrated into the interactive learning platform Artemis that offers personalized, context-aware assistance in large-scale educational settings. Iris supports computer science students by guiding them through programming exercises and is designed to act as a tutor in a didactically meaningful way. Its calibrated assistance avoids revealing complete solutions, offering subtle hints or counter-questions to foster independent problem-solving skills. For each question, it issues multiple prompts in a Chain-of-Thought to GPT-3.5-Turbo. The prompts include a tutor role description and examples of meaningful answers through few-shot learning. Iris employs contextual awareness by accessing the problem statement, student code, and automated feedback to provide tailored advice. An empirical evaluation shows that students perceive Iris as effective because it understands their questions, provides relevant support, and contributes to the learning process. While students consider Iris a valuable tool for programming exercises and homework, they also feel confident solving programming tasks in computer-based exams without Iris. The findings underscore students' appreciation for Iris' immediate and personalized support, though students predominantly view it as a complement to, rather than a replacement for, human tutors. Nevertheless, Iris creates a space for students to ask questions without being judged by others.

Feasible solutions are crucial for Integer Programming (IP) since they can substantially speed up the solving process. In many applications, similar IP instances often exhibit similar structures and shared solution distributions, which can be potentially modeled by deep learning methods. Unfortunately, existing deep-learning-based algorithms, such as Neural Diving and Predict-and-search framework, are limited to generating only partial feasible solutions, and they must rely on solvers like SCIP and Gurobi to complete the solutions for a given IP problem. In this paper, we propose a novel framework that generates complete feasible solutions end-to-end. Our framework leverages contrastive learning to characterize the relationship between IP instances and solutions, and learns latent embeddings for both IP instances and their solutions. Further, the framework employs diffusion models to learn the distribution of solution embeddings conditioned on IP representations, with a dedicated guided sampling strategy that accounts for both constraints and objectives. We empirically evaluate our framework on four typical datasets of IP problems, and show that it effectively generates complete feasible solutions with a high probability (> 89.7 \%) without the reliance of Solvers and the quality of solutions is comparable to the best heuristic solutions from Gurobi. Furthermore, by integrating our method's sampled partial solutions with the CompleteSol heuristic from SCIP, the resulting feasible solutions outperform those from state-of-the-art methods across all datasets, exhibiting a 3.7 to 33.7\% improvement in the gap to optimal values, and maintaining a feasible ratio of over 99.7\% for all datasets.

In Machine Learning (ML)-enabled autonomous systems (MLASs), it is essential to identify the hazard boundary of ML Components (MLCs) in the MLAS under analysis. Given that such boundary captures the conditions in terms of MLC behavior and system context that can lead to hazards, it can then be used to, for example, build a safety monitor that can take any predefined fallback mechanisms at runtime when reaching the hazard boundary. However, determining such hazard boundary for an ML component is challenging. This is due to the problem space combining system contexts (i.e., scenarios) and MLC behaviors (i.e., inputs and outputs) being far too large for exhaustive exploration and even to handle using conventional metaheuristics, such as genetic algorithms. Additionally, the high computational cost of simulations required to determine any MLAS safety violations makes the problem even more challenging. Furthermore, it is unrealistic to consider a region in the problem space deterministically safe or unsafe due to the uncontrollable parameters in simulations and the non-linear behaviors of ML models (e.g., deep neural networks) in the MLAS under analysis. To address the challenges, we propose MLCSHE (ML Component Safety Hazard Envelope), a novel method based on a Cooperative Co-Evolutionary Algorithm (CCEA), which aims to tackle a high-dimensional problem by decomposing it into two lower-dimensional search subproblems. Moreover, we take a probabilistic view of safe and unsafe regions and define a novel fitness function to measure the distance from the probabilistic hazard boundary and thus drive the search effectively. We evaluate the effectiveness and efficiency of MLCSHE on a complex Autonomous Vehicle (AV) case study. Our evaluation results show that MLCSHE is significantly more effective and efficient compared to a standard genetic algorithm and random search.

Various parameters affect the performance of students in online coding competitions. Students' behavior, approach, emotions, and problem difficulty levels significantly impact their performance in online coding competitions. We have organized two coding competitions to understand the effects of the above parameters. We have done the online survey at the end of each coding competition, and it contains questions related to the behavior, approach, and emotions of students during online coding competitions. Students are evaluated based on the time and status of the submissions. We have carried out a detailed analysis to address the impact of students' approach, behavior, and emotions on the learning process in online coding competitions. Two difficulty levels are proposed based on the time and status of submissions. The impact of difficulty levels on machine learning-based performance prediction is presented in this research work. Based on time, the coding solution submissions have two classes "Less than 15 minutes" and "More than 15 minutes". There are three classes, "Complete solution", "Partial solution", and "Not submitted at all," based on the submission status. The appropriate approaches are found for both the coding competitions to submit the solution within 15 minutes. Machine learning classifiers are trained and evaluated for the above classification problems. The impacts of mood, emotions, and difficulty levels on the learning process are also assessed by comparing the results of machine learning models for both coding competitions.

Data science projects vary in scope and complexity. Sometimes, the project could be as simple as producing summary statistics, charts, and visualizations. It could also involve building a regression model, classification model, or forecasting using a time-dependent dataset. The project could also be very complex and difficult, with no clear guidance as to the specific type of model to use. In this case, it is the task of the data science aspirant or professional to come up with a model that best suitable for addressing project goals and objectives.

A linear algorithm is described for solving the n-Queens Completion problem for an arbitrary composition of k queens, consistently distributed on a chessboard of size n x n. Two important rules are used in the algorithm: a) the rule of sequential risk elimination for the entire system as a whole; b) the rule of formation of minimal damage in the given selection conditions. For any composition of k queens (1<= k