Search

Modern time-series forecasting models often fail to make full use of rich unstructured information about the time series themselves. This lack of proper conditioning can lead to obvious model failures; for example, models may be unaware of the details of a particular product, and hence fail to anticipate seasonal surges in customer demand in the lead up to major exogenous events like holidays for clearly relevant products. To address this shortcoming, this paper introduces a novel forecast post-processor -- which we call LLMForecaster -- that fine-tunes large language models (LLMs) to incorporate unstructured semantic and contextual information and historical data to improve the forecasts from an existing demand forecasting pipeline. In an industry-scale retail application, we demonstrate that our technique yields statistically significantly forecast improvements across several sets of products subject to holiday-driven demand surges., Comment: Presented at NeurIPS Time Series in the Age of Large Models (2024)

This paper presents the latest version of the Global AI Vibrancy Tool (GVT), an interactive suite of visualizations designed to facilitate the comparison of AI vibrancy across 36 countries, using 42 indicators organized into 8 pillars. The tool offers customizable features that allow users to conduct in-depth country-level comparisons and longitudinal analyses of AI-related metrics, all based on publicly available data. By providing a transparent assessment of national progress in AI, it serves the diverse needs of policymakers, industry leaders, researchers, and the general public. Using weights for indicators and pillars developed by AI Index's panel of experts and combined into an index, the Global AI Vibrancy Ranking for 2023 places the United States first by a significant margin, followed by China and the United Kingdom. The ranking also highlights the rise of smaller nations such as Singapore when evaluated on both absolute and per capita bases. The tool offers three sub-indices for evaluating Global AI Vibrancy along different dimensions: the Innovation Index, the Economic Competitiveness Index, and the Policy, Governance, and Public Engagement Index.

Objective speech quality measures are typically used to assess speech enhancement algorithms, but it has been shown that they are sub-optimal as learning objectives because they do not always align well with human subjective ratings. This misalignment often results in noticeable distortions and artifacts that cause speech enhancement to be ineffective. To address these issues, we propose a reinforcement learning from human feedback (RLHF) framework to fine-tune an existing speech enhancement approach by optimizing performance using a mean-opinion score (MOS)-based reward model. Our results show that the RLHF-finetuned model has the best performance across different benchmarks for both objective and MOS-based speech quality assessment metrics on the Voicebank+DEMAND dataset. Through ablation studies, we show that both policy gradient loss and supervised MSE loss are important for balanced optimization across the different metrics., Comment: Submitted to ICASSP 2025

In this paper, we introduce a novel approach for bounding the cumulant generating function (CGF) of a Dirichlet process (DP) $X \sim \text{DP}(\alpha \nu_0)$, using superadditivity. In particular, our key technical contribution is the demonstration of the superadditivity of $\alpha \mapsto \log \mathbb{E}_{X \sim \text{DP}(\alpha \nu_0)}[\exp( \mathbb{E}_X[\alpha f])]$, where $\mathbb{E}_X[f] = \int f dX$. This result, combined with Fekete's lemma and Varadhan's integral lemma, converts the known asymptotic large deviation principle into a practical upper bound on the CGF $ \log\mathbb{E}_{X\sim \text{DP}(\alpha\nu_0)}{\exp(\mathbb{E}_{X}{[f]})} $ for any $\alpha > 0$. The bound is given by the convex conjugate of the scaled reversed Kullback-Leibler divergence $\alpha\mathrm{KL}(\nu_0\Vert \cdot)$. This new bound provides particularly effective confidence regions for sums of independent DPs, making it applicable across various fields.

We delve into the challenges of accurately estimating 3D human pose and shape in video surveillance scenarios. Beginning with the advocacy for metrics like W-MPJPE and W-PVE, which omit the (Procrustes) realignment step, to improve model evaluation, we then introduce RotAvat. This technique aims to enhance these metrics by refining the alignment of 3D meshes with the ground plane. Through qualitative comparisons, we demonstrate RotAvat's effectiveness in addressing the limitations of existing aproaches.

Sampling-based motion planners (SBMPs) are effective for planning with complex kinodynamic constraints in high-dimensional spaces, but they still struggle to achieve real-time performance, which is mainly due to their serial computation design. We present Kinodynamic Parallel Accelerated eXpansion (Kino-PAX), a novel highly parallel kinodynamic SBMP designed for parallel devices such as GPUs. Kino-PAX grows a tree of trajectory segments directly in parallel. Our key insight is how to decompose the iterative tree growth process into three massively parallel subroutines. Kino-PAX is designed to align with the parallel device execution hierarchies, through ensuring that threads are largely independent, share equal workloads, and take advantage of low-latency resources while minimizing high-latency data transfers and process synchronization. This design results in a very efficient GPU implementation. We prove that Kino-PAX is probabilistically complete and analyze its scalability with compute hardware improvements. Empirical evaluations demonstrate solutions in the order of 10 ms on a desktop GPU and in the order of 100 ms on an embedded GPU, representing up to 1000 times improvement compared to coarse-grained CPU parallelization of state-of-the-art sequential algorithms over a range of complex environments and systems., Comment: Preprint Under Review

The deliberative potential of online platforms has been widely examined but the impact of reflection time on the quality of deliberation remains under-explored. This paper presents two user studies involving 100 and 72 participants respectively, to investigate the impact of reflection time on the quality of deliberation in minute-scale deliberations. In the first study, we identified an optimal reflection time for composing short opinion comments. In the second study, we introduced four distinct interface-based time nudges aimed at encouraging reflection near the optimal time. While these nudges may not improve the quality of deliberation, they effectively prolonged reflection periods. Additionally, we observed mixed effects on users' experience, influenced by the nature of the time nudges. Our findings suggest that reflection time is crucial, particularly for users who typically deliberate below the optimal reflection threshold.

Large Multimodal Models (LMMs) excel at comprehending human instructions and demonstrate remarkable results across a broad spectrum of tasks. Reinforcement Learning from Human Feedback (RLHF) and AI Feedback (RLAIF) further refine LLMs by aligning them with specific preferences. These methods primarily use ranking-based feedback for entire generations. With advanced AI models (Teacher), such as GPT-4 and Claude 3 Opus, we can request various types of detailed feedback that are expensive for humans to provide. We propose a two-stage algorithm ARES that Alternates REinforcement Learning (RL) and Supervised Fine-Tuning (SFT). First, we request the Teacher to score how much each sentence contributes to solving the problem in a Chain-of-Thought (CoT). This sentence-level feedback allows us to consider individual valuable segments, providing more granular rewards for the RL procedure. Second, we ask the Teacher to correct the wrong reasoning after the RL stage. The RL procedure requires massive efforts for hyperparameter tuning and often generates errors like repetitive words and incomplete sentences. With the correction feedback, we stabilize the RL fine-tuned model through SFT. We conduct experiments on multi-model dataset ScienceQA and A-OKVQA to demonstrate the effectiveness of our proposal. ARES rationale reasoning achieves around 70% win rate against baseline models judged by GPT-4o. Additionally, we observe that the improved rationale reasoning leads to a 2.5% increase in inference answer accuracy on average for the multi-modal datasets., Comment: Accepted to EMNLP 2024

The 2024 Index is our most comprehensive to date and arrives at an important moment when AI's influence on society has never been more pronounced. This year, we have broadened our scope to more extensively cover essential trends such as technical advancements in AI, public perceptions of the technology, and the geopolitical dynamics surrounding its development. Featuring more original data than ever before, this edition introduces new estimates on AI training costs, detailed analyses of the responsible AI landscape, and an entirely new chapter dedicated to AI's impact on science and medicine. The AI Index report tracks, collates, distills, and visualizes data related to artificial intelligence (AI). Our mission is to provide unbiased, rigorously vetted, broadly sourced data in order for policymakers, researchers, executives, journalists, and the general public to develop a more thorough and nuanced understanding of the complex field of AI. The AI Index is recognized globally as one of the most credible and authoritative sources for data and insights on artificial intelligence. Previous editions have been cited in major newspapers, including the The New York Times, Bloomberg, and The Guardian, have amassed hundreds of academic citations, and been referenced by high-level policymakers in the United States, the United Kingdom, and the European Union, among other places. This year's edition surpasses all previous ones in size, scale, and scope, reflecting the growing significance that AI is coming to hold in all of our lives.

Reinforcement learning (RL) training is inherently unstable due to factors such as moving targets and high gradient variance. Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF) can introduce additional difficulty. Differing preferences can complicate the alignment process, and prediction errors in a trained reward model can become more severe as the LLM generates unseen outputs. To enhance training robustness, RL has adopted techniques from supervised learning, such as ensembles and layer normalization. In this work, we improve the stability of RL training by adapting the reverse cross entropy (RCE) from supervised learning for noisy data to define a symmetric RL loss. We demonstrate performance improvements across various tasks and scales. We conduct experiments in discrete action tasks (Atari games) and continuous action space tasks (MuJoCo benchmark and Box2D) using Symmetric A2C (SA2C) and Symmetric PPO (SPPO), with and without added noise with especially notable performance in SPPO across different hyperparameters. Furthermore, we validate the benefits of the symmetric RL loss when using SPPO for large language models through improved performance in RLHF tasks, such as IMDB positive sentiment sentiment and TL;DR summarization tasks.

Recent advancements in generative models have sparked a significant interest within the machine learning community. Particularly, diffusion models have demonstrated remarkable capabilities in synthesizing images and speech. Studies such as those by Lee et al. (2023), Black et al. (2023), Wang et al. (2023), and Fan et al. (2024) illustrate that Reinforcement Learning with Human Feedback (RLHF) can enhance diffusion models for image synthesis. However, due to architectural differences between these models and those employed in speech synthesis, it remains uncertain whether RLHF could similarly benefit speech synthesis models. In this paper, we explore the practical application of RLHF to diffusion-based text-to-speech synthesis, leveraging the mean opinion score (MOS) as predicted by UTokyo-SaruLab MOS prediction system (Saeki et al., 2022) as a proxy loss. We introduce diffusion model loss-guided RL policy optimization (DLPO) and compare it against other RLHF approaches, employing the NISQA speech quality and naturalness assessment model (Mittag et al., 2021) and human preference experiments for further evaluation. Our results show that RLHF can enhance diffusion-based text-to-speech synthesis models, and, moreover, DLPO can better improve diffusion models in generating natural and high quality speech audios.

There is increasing interest in the adoption of LLMs in HCI research. However, LLMs may often be regarded as a panacea because of their powerful capabilities with an accompanying oversight on whether they are suitable for their intended tasks. We contend that LLMs should be adopted in a critical manner following rigorous evaluation. Accordingly, we present the evaluation of an LLM in identifying logical fallacies that will form part of a digital misinformation intervention. By comparing to a labeled dataset, we found that GPT-4 achieves an accuracy of 0.79, and for our intended use case that excludes invalid or unidentified instances, an accuracy of 0.90. This gives us the confidence to proceed with the application of the LLM while keeping in mind the areas where it still falls short. The paper describes our evaluation approach, results and reflections on the use of the LLM for our intended task.

With ChatGPT's release, conversational prompting has become the most popular form of human-LLM interaction. However, its effectiveness is limited for more complex tasks involving reasoning, creativity, and iteration. Through a systematic analysis of HCI papers published since 2021, we identified four key phases in the human-LLM interaction flow - planning, facilitating, iterating, and testing - to precisely understand the dynamics of this process. Additionally, we have developed a taxonomy of four primary interaction modes: Mode 1: Standard Prompting, Mode 2: User Interface, Mode 3: Context-based, and Mode 4: Agent Facilitator. This taxonomy was further enriched using the "5W1H" guideline method, which involved a detailed examination of definitions, participant roles (Who), the phases that happened (When), human objectives and LLM abilities (What), and the mechanics of each interaction mode (How). We anticipate this taxonomy will contribute to the future design and evaluation of human-LLM interaction., Comment: 11 pages, 4 figures, 3 tables. Accepted at CHI Late-Breaking Work 2024

Generative artificial intelligence (GenAI) can rapidly produce large and diverse volumes of content. This lends to it a quality of creativity which can be empowering in the early stages of design. In seeking to understand how creative ways to address practical issues can be conceived between humans and GenAI, we conducted a rapid ideation workshop with 21 participants where they used a large language model (LLM) to brainstorm potential solutions and evaluate them. We found that the LLM produced a greater variety of ideas that were of high quality, though not necessarily of higher quality than human-generated ideas. Participants typically prompted in a straightforward manner with concise instructions. We also observed two collaborative dynamics with the LLM fulfilling a consulting role or an assisting role depending on the goals of the users. Notably, we observed an atypical anti-collaboration dynamic where participants used an antagonistic approach to prompt the LLM.

With fact-checking by professionals being difficult to scale on social media, algorithmic techniques have been considered. However, it is uncertain how the public may react to labels by automated fact-checkers. In this study, we investigate the use of automated warning labels derived from misinformation detection literature and investigate their effects on three forms of post engagement. Focusing on political posts, we also consider how partisanship affects engagement. In a two-phases within-subjects experiment with 200 participants, we found that the generic warnings suppressed intents to comment on and share posts, but not on the intent to like them. Furthermore, when different reasons for the labels were provided, their effects on post engagement were inconsistent, suggesting that the reasons could have undesirably motivated engagement instead. Partisanship effects were observed across the labels with higher engagement for politically congruent posts. We discuss the implications on the design and use of automated warning labels.

In Singapore, there has been a rise in misinformation on mobile instant messaging services (MIMS). MIMS support both small peer-to-peer networks and large groups. Misinformation in the former may spread due to recipients' trust in the sender while in the latter, misinformation can directly reach a wide audience. The encryption of MIMS makes it difficult to address misinformation directly. As such, chatbots have become an alternative solution where users can disclose their chat content directly to fact checking services. To understand how effective fact checking chatbots are as an intervention and how trust in three different fact checkers (i.e., Government, News Outlets, and Artificial Intelligence) may affect this trust, we conducted a within-subjects experiment with 527 Singapore residents. We found mixed results for the fact checkers but support for the chatbot intervention overall. We also found a striking contradiction between participants' trust in the fact checkers and their behaviour towards them. Specifically, those who reported a high level of trust in the government performed worse and tended to follow the fact checking tool less when it was endorsed by the government.

This paper introduces a vision for Quantum Software Development lifecycle, proposing a hybrid full-stack iterative model that integrates quantum and classical computing. Addressing the current challenges in Quantum Computing (QC) such as the need for integrating diverse programming languages and managing the complexities of quantum-classical systems, this model is rooted in the principles of DevOps and continuous software engineering. It presents a comprehensive lifecycle for quantum software development, encompassing quantum-agnostic coding, testing, deployment, cloud computing services, orchestration, translation, execution, and interpretation phases. Each phase is designed to accommodate the unique demands of QC, enabling traditional software developers to engage with QC environments without needing in-depth QC expertise. The paper presents a detailed implementation roadmap, utilizing a range of existing tools and frameworks, thereby making quantum software development more accessible and efficient. The proposed model not only addresses current challenges in quantum software development but also makes a substantial contribution to the field of Quantum Software Engineering (QSE). By proposing a structured and accessible model, it sets the stage for further advancements and research in QSE, enhancing its practicality and relevance in a wide range of applications.

Owing to the increase in 'certified' phishing websites, there is a steady increase in the number of phishing cases and general susceptibility to phishing. Trust mechanisms (e.g., HTTPS Lock Indicators, SSL Certificates) that help differentiate genuine and phishing websites should therefore be evaluated for their effectiveness in preventing vulnerable users from accessing phishing websites. In this article, we present a study involving 18 adults (male-6; female-12) and 12 older adults (male-4; female-8) to understand the usability of current trust mechanisms and preferred modalities in a conceptualized mechanism. In the first part of the study, using Chrome browser on Android, we asked the participants to browse a banking website and a government website for digital particulars. We asked them to identify which one of the two was a phishing website, rate the usability of both websites and provide qualitative feedback on the trust mechanisms. In the second part, we conceptualized an alternative trust mechanism, which allows seeking social, community and AI-based support to make website trust-related decisions. Herein, we asked the participants as to which modality (social, community or AI) they prefer to seek support from and why it is preferred. Using the current trust mechanisms, none of the participants were able to identify the phishing website. As the participants rated the current mechanisms poorly in terms of usability, they expressed various difficulties that largely did not differ between adults and older adults. In the conceptualized mechanism, we observed a notable difference in the preferred modalities, in that, older adults primarily preferred social support. In addition to these overall findings, specific observations suggest that future trust mechanisms should not only consider age-specific needs but also incorporate substantial improvement in terms of usability., Comment: This version was last submitted to EuroUSEC 2023 - European Symposium on Usable Security. It was later invited for poster submission at the same conference

The deliberative potential of online platforms has been widely examined. However, little is known about how various interface-based reflection nudges impact the quality of deliberation. This paper presents two user studies with 12 and 120 participants, respectively, to investigate the impacts of different reflective nudges on the quality of deliberation. In the first study, we examined five distinct reflective nudges: persona, temporal prompts, analogies and metaphors, cultural prompts and storytelling. Persona, temporal prompts, and storytelling emerged as the preferred nudges for implementation on online deliberation platforms. In the second study, we assess the impacts of these preferred reflectors more thoroughly. Results revealed a significant positive impact of these reflectors on deliberative quality. Specifically, persona promotes a deliberative environment for balanced and opinionated viewpoints while temporal prompts promote more individualised viewpoints. Our findings suggest that the choice of reflectors can significantly influence the dynamics and shape the nature of online discussions.

We study reinforcement learning in the presence of an unknown reward perturbation. Existing methodologies for this problem make strong assumptions including reward smoothness, known perturbations, and/or perturbations that do not modify the optimal policy. We study the case of unknown arbitrary perturbations that discretize and shuffle reward space, but have the property that the true reward belongs to the most frequently observed class after perturbation. This class of perturbations generalizes existing classes (and, in the limit, all continuous bounded perturbations) and defeats existing methods. We introduce an adaptive distributional reward critic and show theoretically that it can recover the true rewards under technical conditions. Under the targeted perturbation in discrete and continuous control tasks, we win/tie the highest return in 40/57 settings (compared to 16/57 for the best baseline). Even under the untargeted perturbation, we still win an edge over the baseline designed especially for that setting.

In recent years, there has been growing interest in developing robust machine learning (ML) models that can withstand adversarial attacks, including one of the most widely adopted, efficient, and interpretable ML algorithms-decision trees (DTs). This paper proposes a novel coevolutionary algorithm (CoEvoRDT) designed to create robust DTs capable of handling noisy high-dimensional data in adversarial contexts. Motivated by the limitations of traditional DT algorithms, we leverage adaptive coevolution to allow DTs to evolve and learn from interactions with perturbed input data. CoEvoRDT alternately evolves competing populations of DTs and perturbed features, enabling construction of DTs with desired properties. CoEvoRDT is easily adaptable to various target metrics, allowing the use of tailored robustness criteria such as minimax regret. Furthermore, CoEvoRDT has potential to improve the results of other state-of-the-art methods by incorporating their outcomes (DTs they produce) into the initial population and optimize them in the process of coevolution. Inspired by the game theory, CoEvoRDT utilizes mixed Nash equilibrium to enhance convergence. The method is tested on 20 popular datasets and shows superior performance compared to 4 state-of-the-art algorithms. It outperformed all competing methods on 13 datasets with adversarial accuracy metrics, and on all 20 considered datasets with minimax regret. Strong experimental results and flexibility in choosing the error measure make CoEvoRDT a promising approach for constructing robust DTs in real-world applications.

In this paper, we introduce Randomized Q-learning (RandQL), a novel randomized model-free algorithm for regret minimization in episodic Markov Decision Processes (MDPs). To the best of our knowledge, RandQL is the first tractable model-free posterior sampling-based algorithm. We analyze the performance of RandQL in both tabular and non-tabular metric space settings. In tabular MDPs, RandQL achieves a regret bound of order $\widetilde{\mathcal{O}}(\sqrt{H^{5}SAT})$, where $H$ is the planning horizon, $S$ is the number of states, $A$ is the number of actions, and $T$ is the number of episodes. For a metric state-action space, RandQL enjoys a regret bound of order $\widetilde{\mathcal{O}}(H^{5/2} T^{(d_z+1)/(d_z+2)})$, where $d_z$ denotes the zooming dimension. Notably, RandQL achieves optimistic exploration without using bonuses, relying instead on a novel idea of learning rate randomization. Our empirical study shows that RandQL outperforms existing approaches on baseline exploration environments., Comment: NeurIPS-2023

Incorporating expert demonstrations has empirically helped to improve the sample efficiency of reinforcement learning (RL). This paper quantifies theoretically to what extent this extra information reduces RL's sample complexity. In particular, we study the demonstration-regularized reinforcement learning that leverages the expert demonstrations by KL-regularization for a policy learned by behavior cloning. Our findings reveal that using $N^{\mathrm{E}}$ expert demonstrations enables the identification of an optimal policy at a sample complexity of order $\widetilde{O}(\mathrm{Poly}(S,A,H)/(\varepsilon^2 N^{\mathrm{E}}))$ in finite and $\widetilde{O}(\mathrm{Poly}(d,H)/(\varepsilon^2 N^{\mathrm{E}}))$ in linear Markov decision processes, where $\varepsilon$ is the target precision, $H$ the horizon, $A$ the number of action, $S$ the number of states in the finite case and $d$ the dimension of the feature space in the linear case. As a by-product, we provide tight convergence guarantees for the behaviour cloning procedure under general assumptions on the policy classes. Additionally, we establish that demonstration-regularized methods are provably efficient for reinforcement learning from human feedback (RLHF). In this respect, we provide theoretical evidence showing the benefits of KL-regularization for RLHF in tabular and linear MDPs. Interestingly, we avoid pessimism injection by employing computationally feasible regularization to handle reward estimation uncertainty, thus setting our approach apart from the prior works., Comment: This revision fixes an error due to use of some incorrect results (Lemma 32, Corollary 11 by Talebi & Maillard, 2018) in the proof of Theorem 8. The condition for the RLHF results have slightly changed

Welcome to the sixth edition of the AI Index Report. This year, the report introduces more original data than any previous edition, including a new chapter on AI public opinion, a more thorough technical performance chapter, original analysis about large language and multimodal models, detailed trends in global AI legislation records, a study of the environmental impact of AI systems, and more. The AI Index Report tracks, collates, distills, and visualizes data related to artificial intelligence. Our mission is to provide unbiased, rigorously vetted, broadly sourced data in order for policymakers, researchers, executives, journalists, and the general public to develop a more thorough and nuanced understanding of the complex field of AI. The report aims to be the world's most credible and authoritative source for data and insights about AI.

While AI shows promise for enhancing the efficiency of qualitative analysis, the unique human-AI interaction resulting from varied coding strategies makes it challenging to develop a trustworthy AI-assisted qualitative coding system (AIQCs) that supports coding tasks effectively. We bridge this gap by exploring the impact of varying coding strategies on user trust and reliance on AI. We conducted a mixed-methods split-plot 3x3 study, involving 30 participants, and a follow-up study with 6 participants, exploring varying text selection and code length in the use of our AIQCs system for qualitative analysis. Our results indicate that qualitative open coding should be conceptualized as a series of distinct subtasks, each with differing levels of complexity, and therefore, should be given tailored design considerations. We further observed a discrepancy between perceived and behavioral measures, and emphasized the potential challenges of under- and over-reliance on AIQCs systems. Additional design implications were also proposed for consideration., Comment: 27 pages with references, 9 figures, 5 tables

The massive volume of online information along with the issue of misinformation has spurred active research in the automation of fact-checking. Like fact-checking by human experts, it is not enough for an automated fact-checker to just be accurate, but also be able to inform and convince the user of the validity of its predictions. This becomes viable with explainable artificial intelligence (XAI). In this work, we conduct a study of XAI fact-checkers involving 180 participants to determine how users' actions towards news and their attitudes towards explanations are affected by the XAI. Our results suggest that XAI has limited effects on users' agreement with the veracity prediction of the automated fact-checker and on their intent to share news. However, XAI nudges users towards forming uniform judgments of news veracity, thereby signaling their reliance on the explanations. We also found polarizing preferences towards XAI and raise several design considerations on them.

In this white paper, we synthesize key points made during presentations and discussions from the AI-Assisted Decision Making for Conservation workshop, hosted by the Center for Research on Computation and Society at Harvard University on October 20-21, 2022. We identify key open research questions in resource allocation, planning, and interventions for biodiversity conservation, highlighting conservation challenges that not only require AI solutions, but also require novel methodological advances. In addition to providing a summary of the workshop talks and discussions, we hope this document serves as a call-to-action to orient the expansion of algorithmic decision-making approaches to prioritize real-world conservation challenges, through collaborative efforts of ecologists, conservation decision-makers, and AI researchers., Comment: Co-authored by participants from the October 2022 workshop: https://crcs.seas.harvard.edu/conservation-workshop

Predict-then-Optimize is a framework for using machine learning to perform decision-making under uncertainty. The central research question it asks is, "How can the structure of a decision-making task be used to tailor ML models for that specific task?" To this end, recent work has proposed learning task-specific loss functions that capture this underlying structure. However, current approaches make restrictive assumptions about the form of these losses and their impact on ML model behavior. These assumptions both lead to approaches with high computational cost, and when they are violated in practice, poor performance. In this paper, we propose solutions to these issues, avoiding the aforementioned assumptions and utilizing the ML model's features to increase the sample efficiency of learning loss functions. We empirically show that our method achieves state-of-the-art results in four domains from the literature, often requiring an order of magnitude fewer samples than comparable methods from past work. Moreover, our approach outperforms the best existing method by nearly 200% when the localness assumption is broken., Comment: 10 pages, 2 figures

Collaborative Qualitative Analysis (CQA) can enhance qualitative analysis rigor and depth by incorporating varied viewpoints. Nevertheless, ensuring a rigorous CQA procedure itself can be both demanding and costly. To lower this bar, we take a theoretical perspective to design the CollabCoder workflow, that integrates Large Language Models (LLMs) into key inductive CQA stages: independent open coding, iterative discussions, and final codebook creation. In the open coding phase, CollabCoder offers AI-generated code suggestions and records decision-making data. During discussions, it promotes mutual understanding by sharing this data within the coding team and using quantitative metrics to identify coding (dis)agreements, aiding in consensus-building. In the code grouping stage, CollabCoder provides primary code group suggestions, lightening the cognitive load of finalizing the codebook. A 16-user evaluation confirmed the effectiveness of CollabCoder, demonstrating its advantages over existing software and providing empirical insights into the role of LLMs in the CQA practice., Comment: Will be published at the ACM CHI Conference on Human Factors in Computing Systems (CHI'24)

While AI-assisted individual qualitative analysis has been substantially studied, AI-assisted collaborative qualitative analysis (CQA)-a process that involves multiple researchers working together to interpret data-remains relatively unexplored. After identifying CQA practices and design opportunities through formative interviews, we designed and implemented CoAIcoder, a tool leveraging AI to enhance human-to-human collaboration within CQA through four distinct collaboration methods. With a between-subject design, we evaluated CoAIcoder with 32 pairs of CQA-trained participants across common CQA phases under each collaboration method. Our findings suggest that while using a shared AI model as a mediator among coders could improve CQA efficiency and foster agreement more quickly in the early coding stage, it might affect the final code diversity. We also emphasize the need to consider the independence level when using AI to assist human-to-human collaboration in various CQA scenarios. Lastly, we suggest design implications for future AI-assisted CQA systems., Comment: Will appear on ACM Transactions on Computer-Human Interaction (TOCHI)

We address the challenge of exploration in reinforcement learning (RL) when the agent operates in an unknown environment with sparse or no rewards. In this work, we study the maximum entropy exploration problem of two different types. The first type is visitation entropy maximization previously considered by Hazan et al.(2019) in the discounted setting. For this type of exploration, we propose a game-theoretic algorithm that has $\widetilde{\mathcal{O}}(H^3S^2A/\varepsilon^2)$ sample complexity thus improving the $\varepsilon$-dependence upon existing results, where $S$ is a number of states, $A$ is a number of actions, $H$ is an episode length, and $\varepsilon$ is a desired accuracy. The second type of entropy we study is the trajectory entropy. This objective function is closely related to the entropy-regularized MDPs, and we propose a simple algorithm that has a sample complexity of order $\widetilde{\mathcal{O}}(\mathrm{poly}(S,A,H)/\varepsilon)$. Interestingly, it is the first theoretical result in RL literature that establishes the potential statistical advantage of regularized MDPs for exploration. Finally, we apply developed regularization techniques to reduce sample complexity of visitation entropy maximization to $\widetilde{\mathcal{O}}(H^2SA/\varepsilon^2)$, yielding a statistical separation between maximum entropy exploration and reward-free exploration., Comment: ICML-2023

We study the Combinatorial Thompson Sampling policy (CTS) for combinatorial multi-armed bandit problems (CMAB), within an approximation regret setting. Although CTS has attracted a lot of interest, it has a drawback that other usual CMAB policies do not have when considering non-exact oracles: for some oracles, CTS has a poor approximation regret (scaling linearly with the time horizon $T$) [Wang and Chen, 2018]. A study is then necessary to discriminate the oracles on which CTS could learn. This study was started by Kong et al. [2021]: they gave the first approximation regret analysis of CTS for the greedy oracle, obtaining an upper bound of order $\mathcal{O}(\log(T)/\Delta^2)$, where $\Delta$ is some minimal reward gap. In this paper, our objective is to push this study further than the simple case of the greedy oracle. We provide the first $\mathcal{O}(\log(T)/\Delta)$ approximation regret upper bound for CTS, obtained under a specific condition on the approximation oracle, allowing a reduction to the exact oracle analysis. We thus term this condition REDUCE2EXACT, and observe that it is satisfied in many concrete examples. Moreover, it can be extended to the probabilistically triggered arms setting, thus capturing even more problems, such as online influence maximization., Comment: Neurips 2022