Recommender systems (RS), which are widely deployed across high-stakes domains, are susceptible to biases that can cause large-scale societal impacts. Researchers have proposed methods to measure and mitigate such biases—but translating academic theory into practice is inherently challenging. Through a semi-structured interview study (N=11), we map the RS practitioner workflow within large technology companies, focusing on how technical teams consider fairness internally and in collaboration with legal, data, and fairness teams. We identify key challenges to incorporating fairness into existing RS workflows: defining fairness in RS contexts, balancing multi-stakeholder interests, and navigating dynamic environments. We also identify key organization-wide challenges: making time for fairness work and facilitating cross-team communication. Finally, we offer actionable recommendations for the RS community, including practitioners and HCI researchers.

Junxiong Wang, Ali Mousavi, Omar Attia, Ronak Pradeep, Saloni Potdar, Alexander Rush, Umar Farooq Minhas, Yunyao Li. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Linear RNN architectures, like Mamba, can be competitive with Transformer models in language modeling while having advantageous deployment characteristics. Given the focus on training large-scale Transformer models, we consider the challenge of converting these pretrained models for deployment. We demonstrate that it is feasible to distill large Transformers into linear RNNs by reusing the linear projection weights from attention layers with academic GPU resources. The resulting hybrid model, which incorporates a quarter of the attention layers, achieves performance comparable to the original Transformer in chat benchmarks and outperforms open-source hybrid Mamba models trained from scratch with trillions of tokens in both chat benchmarks and general benchmarks. Moreover, we introduce a hardware-aware speculative decoding algorithm that accelerates the inference speed of Mamba and hybrid models. Overall we show how, with limited computation resources, we can remove many of the original attention layers and generate from the resulting model more efficiently. Our top-performing model, distilled from Llama3-8B-Instruct, achieves a 29.61 length-controlled win rate on AlpacaEval 2 against GPT-4 and 7.35 on MT-Bench, surpassing the best 8B scale instruction-tuned linear RNN model. We also find that the distilled model has natural length extrapolation, showing almost perfect accuracy in the needle-in-a-haystack test at 20x the distillation length. Code and pre-trained checkpoints are open-sourced at [MambaInLlama](https://github.com/jxiw/MambaInLlama) for distillation and [SpeculativeMamba](https://github.com/itsdaniele/speculative\_mamba) for speculative decoding.

Performance of worst-case optimal join algorithms depends on the order in which the join attributes are processed. It is challenging to identify suitable orders prior to query execution due to the huge search space of possible orders and unreliable execution cost estimates in case of data skew or data correlation. We demonstrate ADOPT, a novel query engine that integrates adaptive query processing with a worst-case optimal join algorithm. ADOPT divides query execution into episodes, during which different attribute orders are invoked. With runtime feedback on performance of different attribute orders, ADOPT rapidly approaches near-optimal orders. Moreover, ADOPT uses a unique data structure which keeps track of the processed input data to prevent redundant work across different episodes. It selects attribute orders to try via reinforcement learning, balancing the need for exploring new orders with the desire to exploit promising orders. In experiments, ADOPT outperforms baselines, including commercial and open-source systems utilizing worst-case optimal join algorithms, particularly for complex queries that are difficult to optimize.

The performance of worst-case optimal join algorithms depends on the order in which the join attributes are processed. Selecting good orders before query execution is hard, due to the large space of possible orders and unreliable execution cost estimates in case of data skew or data correlation. We propose ADOPT, a query engine that combines adaptive query processing with a worst-case optimal join algorithm, which uses an order on the join attributes instead of a join order on relations. ADOPT divides query execution into episodes in which different attribute orders are tried. Based on run time feedback on attribute order performance, ADOPT converges quickly to near-optimal orders. It avoids redundant work across different orders via a novel data structure, keeping track of parts of the join input that have been successfully processed. It selects attribute orders to try via reinforcement learning, balancing the need for exploring new orders with the desire to exploit promising orders. In experiments with various data sets and queries, it outperforms baselines, including commercial and open-source systems using worst-case optimal join algorithms, whenever queries become complex and therefore difficult to optimize.

In black-box optimization problems, we aim to maximize an unknown objective function, where the function is only accessible through feedbacks of an evaluation or simulation oracle. In real-life, the feedbacks of such oracles are often noisy and available after some unknown delay that may depend on the computation time of the oracle. Additionally, if the exact evaluations are expensive but coarse approximations are available at a lower cost, the feedbacks can have multi-fidelity. In order to address this problem, we propose a generic extension of hierarchical optimistic tree search (HOO), called ProCrastinated Tree Search (PCTS), that flexibly accommodates a delay and noise-tolerant bandit algorithm. We provide a generic proof technique to quantify regret of PCTS under delayed, noisy, and multi-fidelity feedbacks. Specifically, we derive regret bounds of PCTS enabled with delayed-UCB1 (DUCB1) and delayed-UCB-V (DUCBV) algorithms. Given a horizon T, PCTS retains the regret bound of non-delayed HOO for expected delay of O(log T), and worsens by T^((1-α)/(d+2)) for expected delays of O(T^(1-α)) for α ∈ (0,1]. We experimentally validate on multiple synthetic functions and hyperparameter tuning problems that PCTS outperforms the state-of-the-art black-box optimization methods for feedbacks with different noise levels, delays, and fidelity.

Index tuning aims to find the optimal index configuration for an input workload. It is a resource-intensive task since it requires making multiple expensive "what-if" calls to the query optimizer to estimate the cost of a query given an index configuration without actually building the indexes. In this paper, we study the problem of budget-aware index tuning where the number of what-if calls allowed when searching for the optimal configuration during tuning is constrained. This problem is challenging as it requires addressing the trade-off between investing what-if calls on exploring new configurations versus exploiting a known promising configuration. We formulate budget-aware index tuning as a Markov decision process, and propose a solution based on Monte Carlo tree search, a classic reinforcement learning technology. Experimental evaluation on both standard industry benchmarks and real workloads shows that our solution can significantly outperform alternative budget-aware solutions in terms of the quality of the index configuration.

UDO is a versatile tool for offline tuning of database systems for specific workloads. UDO can consider a variety of tuning choices, reaching from picking transaction code variants over index selections up to database system parameter tuning. UDO uses reinforcement learning to converge to near-optimal configurations, creating and evaluating different configurations via actual query executions (instead of relying on simplifying cost models). To cater to different parameter types, UDO distinguishes heavy parameters (which are expensive to change, e.g. physical design parameters) from light parameters. Specifically for optimizing heavy parameters, UDO uses reinforcement learning algorithms that allow delaying the point at which reward feedback becomes available. This gives us the freedom to optimize the point in time and the order in which different configurations are created and evaluated (by benchmarking a workload sample). UDO uses a cost-based planner to minimize configuration switching overheads. For instance, it aims to amortize the creation of expensive data structures by consecutively evaluating configurations using them. We demonstrate UDO on Postgres as well as MySQL and on TPC-H as well as TPC-C, optimizing a variety of light and heavy parameters concurrently.

UDO is a versatile tool for offline tuning of database systems for specific workloads. UDO can consider a variety of tuning choices, reaching from picking transaction code variants over index selections up to database system parameter tuning. UDO uses reinforcement learning to converge to near-optimal configurations, creating and evaluating different configurations via actual query executions (instead of relying on simplifying cost models). To cater to different parameter types, UDO distinguishes heavy parameters (which are expensive to change, e.g. physical design parameters) from light parameters. Specifically for optimizing heavy parameters, UDO uses reinforcement learning algorithms that allow delaying the point at which the reward feedback becomes available. This gives us the freedom to optimize the point in time and the order in which different configurations are created and evaluated (by benchmarking a workload sample). UDO uses a cost-based planner to minimize reconfiguration overheads. For instance, it aims to amortize the creation of expensive data structures by consecutively evaluating configurations using them. We evaluate UDO on Postgres as well as MySQL and on TPC-H as well as TPC-C, optimizing a variety of light and heavy parameters concurrently.

SkinnerDB uses reinforcement learning for reliable join ordering, exploiting an adaptive processing engine with specialized join algorithms and data structures. It maintains no data statistics and uses no cost or cardinality models. Also, it uses no training workloads nor does it try to link the current query to seemingly similar queries in the past. Instead, it uses reinforcement learning to learn optimal join orders from scratch during the execution of the current query. To that purpose, it divides the execution of a query into many small time slices. Different join orders are tried in different time slices. SkinnerDB merges result tuples generated according to different join orders until a complete query result is obtained. By measuring execution progress per time slice, it identifies promising join orders as execution proceeds. Along with SkinnerDB, we introduce a new quality criterion for query execution strategies. We upper-bound expected execution cost regret, i.e., the expected amount of execution cost wasted due to sub-optimal join order choices. SkinnerDB features multiple execution strategies that are optimized for that criterion. Some of them can be executed on top of existing database systems. For maximal performance, we introduce a customized execution engine, facilitating fast join order switching via specialized multi-way join algorithms and tuple representations. We experimentally compare SkinnerDB’s performance against various baselines, including MonetDB, Postgres, and adaptive processing methods. We consider various benchmarks, including the join order benchmark, TPC-H, and JCC-H, as well as benchmark variants with user-defined functions. Overall, the overheads of reliable join ordering are negligible compared to the performance impact of the occasional, catastrophic join order choice.

SkinnerDB is designed from the ground up for reliable join ordering. It maintains no data statistics and uses no cost or cardinality models. Instead, it uses reinforcement learning to learn optimal join orders on the fly, during the execution of the current query. To that purpose, we divide the execution of a query into many small time slices. Different join orders are tried in different time slices. We merge result tuples generated according to different join orders until a complete result is obtained. By measuring execution progress per time slice, we identify promising join orders as execution proceeds. Along with SkinnerDB, we introduce a new quality criterion for query execution strategies. We compare expected execution cost against execution cost for an optimal join order. SkinnerDB features multiple execution strategies that are optimized for that criterion. Some of them can be executed on top of existing database systems. For maximal performance, we introduce a customized execution engine, facilitating fast join order switching via specialized multi-way join algorithms and tuple representations. We experimentally compare SkinnerDB's performance against various baselines, including MonetDB, Postgres, and adaptive processing methods. We consider various benchmarks, including the join order benchmark and TPC-H variants with user-defined functions. Overall, the overheads of reliable join ordering are negligible compared to the performance impact of the occasional, catastrophic join order choice.

More and more sensitive information of big data are being now stored in the cloud for it offers quality service and convenient management. But the Cloud Server cannot be fully trusted because it may leak information which raises security and privacy concerns of the stored data. Therefore, sensitive data has to be encrypted before shifted to the cloud. It is a great convenience for authorized users to directly retrieve the encrypted data stored in the cloud by keyword search. As we all known, returning all the matching documents is a waste of network bandwidth. Although there are some schemes that support sorting fuzzy multi-keyword search, and they are generally based on edit distance to looking for fuzzy alternatives set, which leads to significantly larger index file size and higher search complexity. Then we rank the search results according to the matching score between keywords and documents. In this paper, we give a novel multi-keyword fuzzy and sortable search scheme. We use n-gram technology to realize fuzzy search and give a novel ranking search scheme based on the fuzzy multi- keyword search. For the matching files may be more than one, there are differences between the original query keywords and the fuzzy keywords. When ranking files, we will not only consider the relevance score between the index keywords and the documents, but also the similarity between the query and keywords. Thus we calculate the comprehensive matching score of the file to the set of query keywords, which is more efficient and accurate.