Orbit: A Framework for Designing and Evaluating Multi-objective Rankers

Machine learning in production needs to balance multiple objectives: This is particularly evident in ranking or recommendation models, where conflicting objectives such as user engagement, satisfaction, diversity, and novelty must be considered at the same time. However, designing multi-objective rankers is inherently a dynamic wicked problem -- there is no single optimal solution, and the needs evolve over time. Effective design requires collaboration between cross-functional teams and careful analysis of a wide range of information. In this work, we introduce Orbit, a conceptual framework for Objective-centric Ranker Building and Iteration. The framework places objectives at the center of the design process, to serve as boundary objects for communication and guide practitioners for design and evaluation. We implement Orbit as an interactive system, which enables stakeholders to interact with objective spaces directly and supports real-time exploration and evaluation of design trade-offs. We evaluate Orbit through a user study involving twelve industry practitioners, showing that it supports efficient design space exploration, leads to more informed decision-making, and enhances awareness of the inherent trade-offs of multiple objectives. Orbit (1) opens up new opportunities of an objective-centric design process for any multi-objective ML models, as well as (2) sheds light on future designs that push practitioners to go beyond a narrow metric-centric or example-centric mindset.

What Is Wrong with My Model? Identifying Systematic Problems with Semantic Data Slicing

Machine learning models make mistakes, yet sometimes it is difficult to identify the systematic problems behind the mistakes. Practitioners engage in various activities, including error analysis, testing, auditing, and red-teaming, to form hypotheses of what can go (or has gone) wrong with their models. To validate these hypotheses, practitioners employ data slicing to identify relevant examples. However, traditional data slicing is limited by available features and programmatic slicing functions. In this work, we propose SemSlicer, a framework that supports semantic data slicing, which identifies a semantically coherent slice, without the need for existing features. SemSlicer uses Large Language Models to annotate datasets and generate slices from any user-defined slicing criteria. We show that SemSlicer generates accurate slices with low cost, allows flexible trade-offs between different design dimensions, reliably identifies under-performing data slices, and helps practitioners identify useful data slices that reflect systematic problems.

Recursive GNNs for Learning Precoding Policies with Size-Generalizability

Graph neural networks (GNNs) have been shown promising in optimizing power allocation and link scheduling with good size generalizability and low training complexity. These merits are important for learning wireless policies under dynamic environments, which partially come from the matched permutation equivariance (PE) properties of the GNNs to the policies to be learned. Nonetheless, it has been noticed in literature that only satisfying the PE property of a precoding policy in multi-antenna systems cannot ensure a GNN for learning precoding to be generalizable to the unseen number of users. Incorporating models with GNNs helps improve size generalizability, which however is only applicable to specific problems, settings, and algorithms. In this paper, we propose a framework of size generalizable GNNs for learning precoding policies that are purely data-driven and can learn wireless policies including but not limited to baseband and hybrid precoding in multi-user multi-antenna systems. To this end, we first find a special structure of each iteration of two numerical algorithms for optimizing precoding, from which we identify the key characteristics of a GNN that affect its size generalizability. Then, we design size-generalizable GNNs that are with these key characteristics and satisfy the PE properties of precoding policies in a recursive manner. Simulation results show that the proposed GNNs can be well-generalized to the number of users for learning baseband and hybrid precoding policies and require much fewer samples than existing counterparts to achieve the same performance.

(Why) Is My Prompt Getting Worse? Rethinking Regression Testing for Evolving LLM APIs

Large Language Models (LLMs) are increasingly integrated into software applications. Downstream application developers often access LLMs through APIs provided as a service. However, LLM APIs are often updated silently and scheduled to be deprecated, forcing users to continuously adapt to evolving models. This can cause performance regression and affect prompt design choices, as evidenced by our case study on toxicity detection. Based on our case study, we emphasize the need for and re-examine the concept of regression testing for evolving LLM APIs. We argue that regression testing LLMs requires fundamental changes to traditional testing approaches, due to different correctness notions, prompting brittleness, and non-determinism in LLM APIs.

Beyond Testers' Biases: Guiding Model Testing with Knowledge Bases using LLMs

Current model testing work has mostly focused on creating test cases. Identifying what to test is a step that is largely ignored and poorly supported. We propose Weaver, an interactive tool that supports requirements elicitation for guiding model testing. Weaver uses large language models to generate knowledge bases and recommends concepts from them interactively, allowing testers to elicit requirements for further testing. Weaver provides rich external knowledge to testers and encourages testers to systematically explore diverse concepts beyond their own biases. In a user study, we show that both NLP experts and non-experts identified more, as well as more diverse concepts worth testing when using Weaver. Collectively, they found more than 200 failing test cases for stance detection with zero-shot ChatGPT. Our case studies further show that Weaver can help practitioners test models in real-world settings, where developers define more nuanced application scenarios (e.g., code understanding and transcript summarization) using LLMs.

Learning Resource Allocation Policy: Vertex-GNN or Edge-GNN?

Graph neural networks (GNNs) update the hidden representations of vertices (called Vertex-GNNs) or hidden representations of edges (called Edge-GNNs) by processing and pooling the information of neighboring vertices and edges and combining to incorporate graph topology. When learning resource allocation policies, GNNs cannot perform well if their expressive power are weak, i.e., if they cannot differentiate all input features such as channel matrices. In this paper, we analyze the expressive power of the Vertex-GNNs and Edge-GNNs for learning three representative wireless policies: link scheduling, power control, and precoding policies. We find that the expressive power of the GNNs depend on the linearity and output dimensions of the processing and combination functions. When linear processors are used, the Vertex-GNNs cannot differentiate all channel matrices due to the loss of channel information, while the Edge-GNNs can. When learning the precoding policy, even the Vertex-GNNs with non-linear processors may not be with strong expressive ability due to the dimension compression. We proceed to provide necessary conditions for the GNNs to well learn the precoding policy. Simulation results validate the analyses and show that the Edge-GNNs can achieve the same performance as the Vertex-GNNs with much lower training and inference time.

LLMs as Workers in Human-Computational Algorithms? Replicating Crowdsourcing Pipelines with LLMs

LLMs have shown promise in replicating human-like behavior in crowdsourcing tasks that were previously thought to be exclusive to human abilities. However, current efforts focus mainly on simple atomic tasks. We explore whether LLMs can replicate more complex crowdsourcing pipelines. We find that modern LLMs can simulate some of crowdworkers' abilities in these "human computation algorithms," but the level of success is variable and influenced by requesters' understanding of LLM capabilities, the specific skills required for sub-tasks, and the optimal interaction modality for performing these sub-tasks. We reflect on human and LLMs' different sensitivities to instructions, stress the importance of enabling human-facing safeguards for LLMs, and discuss the potential of training humans and LLMs with complementary skill sets. Crucially, we show that replicating crowdsourcing pipelines offers a valuable platform to investigate (1) the relative strengths of LLMs on different tasks (by cross-comparing their performances on sub-tasks) and (2) LLMs' potential in complex tasks, where they can complete part of the tasks while leaving others to humans.

Understanding the Usability of AI Programming Assistants

The software engineering community recently has witnessed widespread deployment of AI programming assistants, such as GitHub Copilot. However, in practice, developers do not accept AI programming assistants' initial suggestions at a high frequency. This leaves a number of open questions related to the usability of these tools. To understand developers' practices while using these tools and the important usability challenges they face, we administered a survey to a large population of developers and received responses from a diverse set of 410 developers. Through a mix of qualitative and quantitative analyses, we found that developers are most motivated to use AI programming assistants because they help developers reduce key-strokes, finish programming tasks quickly, and recall syntax, but resonate less with using them to help brainstorm potential solutions. We also found the most important reasons why developers do not use these tools are because these tools do not output code that addresses certain functional or non-functional requirements and because developers have trouble controlling the tool to generate the desired output. Our findings have implications for both creators and users of AI programming assistants, such as designing minimal cognitive effort interactions with these tools to reduce distractions for users while they are programming.

Multidimensional Graph Neural Networks for Wireless Communications

Graph neural networks (GNNs) have been shown promising in improving the efficiency of learning communication policies by leveraging their permutation properties. Nonetheless, existing works design GNNs only for specific wireless policies, lacking a systematical approach for modeling graph and selecting structure. Based on the observation that the mismatched permutation property from the policies and the information loss during the update of hidden representations have large impact on the learning performance and efficiency, in this paper we propose a unified framework to learn permutable wireless policies with multidimensional GNNs. To avoid the information loss, the GNNs update the hidden representations of hyper-edges. To exploit all possible permutations of a policy, we provide a method to identify vertices in a graph. We also investigate the permutability of wireless channels that affects the sample efficiency, and show how to trade off the training, inference, and designing complexities of GNNs. We take precoding in different systems as examples to demonstrate how to apply the framework. Simulation results show that the proposed GNNs can achieve close performance to numerical algorithms, and require much fewer training samples and trainable parameters to achieve the same learning performance as the commonly used convolutional neural networks.

A Model-based GNN for Learning Precoding

Learning precoding policies with neural networks enables low complexity online implementation, robustness to channel impairments, and joint optimization with channel acquisition. However, existing neural networks suffer from high training complexity and poor generalization ability when they are used to learn to optimize precoding for mitigating multi-user interference. This impedes their use in practical systems where the number of users is time-varying. In this paper, we propose a graph neural network (GNN) to learn precoding policies by harnessing both the mathematical model and the property of the policies. We first show that a vanilla GNN cannot well-learn pseudo-inverse of channel matrix when the numbers of antennas and users are large, and is not generalizable to unseen numbers of users. Then, we design a GNN by resorting to the Taylor's expansion of matrix pseudo-inverse, which allows for capturing the importance of the neighbored edges to be aggregated that is crucial for learning precoding policies efficiently. Simulation results show that the proposed GNN can well learn spectral efficient and energy efficient precoding policies in single- and multi-cell multi-user multi-antenna systems with low training complexity, and can be well generalized to the numbers of users.