Md Arif Rahman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Md Arif Rahman has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Shayhan Ameen Chowdhury
Dept. of Computer Science and Engineering (CSE), International Islamic University Chittagong (IIUC), Chittagong-4318, Bangladesh
2 shared publications - Syed Jalaluddin Hashmi
Department of Artificial Intelligence, Kyung Hee University, Yongin-si, Republic of Korea
2 shared publications - Taeyeon Kim
Jashore University of Science and Technology
2 shared publications Young-Koo Lee
2 shared publications- Young-Koo Lee
Department of Artificial Intelligence, Kyung Hee University, Yongin-si, Republic of Korea
2 shared publications - Irfan Ullah
Department of Computer Engineering, Kyung Hee University, Yongin-si, Republic of Korea
1 shared publication
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1538-1548
AdaptSteer: Workload-Adaptive Optimizer Steering via Execution-Derived Preference Supervision
Database query optimizers often select suboptimal execution plans, and effective steering requires identifying configurations that improve query and workload performance without modifying DBMS internals. Existing external steering systems can reason over multiple optimizer configurations, but many rely on online feedback, exploration, or engine-specific steering mechanisms. More recent LLM-based SQL steering approaches avoid deeper optimizer integration, but often depend on narrow supervision and general-purpose embeddings that are costly, not explicitly aligned with execution behavior, and difficult to adapt under workload shifts. We present AdaptSteer, a novel workload-adaptive optimizer-steering framework that learns from execution-derived preference supervision. AdaptSteer converts latencies measured across diverse query--configuration pairs into preference-based contrastive training signals for a compact SQL encoder, without invasive DBMS modifications or external APIs. The resulting representation supports two lightweight decision heads: a calibrated classifier for safety-first binary steering and a pairwise Condorcet ranker that decomposes K-way configuration selection into binary execution-preference comparisons. We further apply Reptile-based meta-learning to the trained encoder, enabling few-shot adaptation to unseen workloads. Evaluations on the JOB and CEB benchmarks show that AdaptSteer reduces total workload latency by 70.1% relative to the PostgreSQL default and achieves 4.4% lower total workload latency than LLMSteer under the same binary protocol, without relying on external API calls. Beyond binary steering, AdaptSteer improves multi-action selection through pairwise preference ranking across the evaluated candidate sets and benefits from multi-configuration supervision across broader action spaces. Under CEB?JOB adaptation, AdaptSteer's meta-learned variant improves AUROC by 4.2 percentage points over the contrastive-only baseline with only 20 target-workload adaptation queries. Together, AdaptSteer unifies execution-derived preference supervision, binary and multi-action steering, and few-shot adaptation in a compact, deployable optimizer-steering framework.
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1538-1548
Open AccessAdaptSteer: Workload-Adaptive Optimizer Steering via Execution-Derived Preference Supervision
Database query optimizers often select suboptimal execution plans, and effective steering requires identifying configurations that improve query and workload performance without modifying DBMS internals. Existing external steering systems can reason over multiple optimizer configurations, but many rely on online feedback, exploration, or engine-specific steering mechanisms. More recent LLM-based SQL steering approaches avoid deeper optimizer integration, but often depend on narrow supervision and general-purpose embeddings that are costly, not explicitly aligned with execution behavior, and difficult to adapt under workload shifts. We present AdaptSteer, a novel workload-adaptive optimizer-steering framework that learns from execution-derived preference supervision. AdaptSteer converts latencies measured across diverse query--configuration pairs into preference-based contrastive training signals for a compact SQL encoder, without invasive DBMS modifications or external APIs. The resulting representation supports two lightweight decision heads: a calibrated classifier for safety-first binary steering and a pairwise Condorcet ranker that decomposes K-way configuration selection into binary execution-preference comparisons. We further apply Reptile-based meta-learning to the trained encoder, enabling few-shot adaptation to unseen workloads. Evaluations on the JOB and CEB benchmarks show that AdaptSteer reduces total workload latency by 70.1% relative to the PostgreSQL default and achieves 4.4% lower total workload latency than LLMSteer under the same binary protocol, without relying on external API calls. Beyond binary steering, AdaptSteer improves multi-action selection through pairwise preference ranking across the evaluated candidate sets and benefits from multi-configuration supervision across broader action spaces. Under CEB?JOB adaptation, AdaptSteer's meta-learned variant improves AUROC by 4.2 percentage points over the contrastive-only baseline with only 20 target-workload adaptation queries. Together, AdaptSteer unifies execution-derived preference supervision, binary and multi-action steering, and few-shot adaptation in a compact, deployable optimizer-steering framework.
- Open Access
- 14 June 2026
- Published by Elsevier in SSRN Electronic Journal
Top-𝐾 Pushdown vs Pull-up: Optimizing Multi-Embedding Query Processing for RAG Workloads in Vector Databases
Retrieval-Augmented Generation (RAG) systems increasingly rely on vector databases to retrieve relevant context using multiple embedding criteria. In such workloads, query processing typically performs independent approximate nearest neighbor (ANN) retrieval per criterion followed by candidate fusion. A key challenge is determining how many candidates to retrieve per criterion: retrieving too many increases execution cost, while retrieving too few risks degrading result quality. In practice, the absence of a principled strategy forces systems to over-retrieve, inflating production costs, or under-retrieve, leading to missed relevant results and reduced user satisfaction. Existing systems address this trade-off using heuristic or fixed retrieval depths, lacking a principled mechanism to balance efficiency and effectiveness. In this paper, we propose an adaptive top-𝐾 pushdown vs pull-up optimization framework for multi- embedding query processing. We formalize query execution as a two-stage pipeline and identify per-criterion retrieval depths as key optimization variables governing both execution cost and retrieval quality. We further introduce an execution-driven cost–quality estimator and formulate a constrained optimization problem solved using linear programming to minimize execution cost while preserving retrieval quality guarantees. Experiments on the MovieLens dataset using pgvector show up to 97× speedup while preserving high quality (98% relevance) within the top 40% pushdown.
- 14 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessTop-𝐾 Pushdown vs Pull-up: Optimizing Multi-Embedding Query Processing for RAG Workloads in Vector Databases
Retrieval-Augmented Generation (RAG) systems increasingly rely on vector databases to retrieve relevant context using multiple embedding criteria. In such workloads, query processing typically performs independent approximate nearest neighbor (ANN) retrieval per criterion followed by candidate fusion. A key challenge is determining how many candidates to retrieve per criterion: retrieving too many increases execution cost, while retrieving too few risks degrading result quality. In practice, the absence of a principled strategy forces systems to over-retrieve, inflating production costs, or under-retrieve, leading to missed relevant results and reduced user satisfaction. Existing systems address this trade-off using heuristic or fixed retrieval depths, lacking a principled mechanism to balance efficiency and effectiveness. In this paper, we propose an adaptive top-𝐾 pushdown vs pull-up optimization framework for multi- embedding query processing. We formalize query execution as a two-stage pipeline and identify per-criterion retrieval depths as key optimization variables governing both execution cost and retrieval quality. We further introduce an execution-driven cost–quality estimator and formulate a constrained optimization problem solved using linear programming to minimize execution cost while preserving retrieval quality guarantees. Experiments on the MovieLens dataset using pgvector show up to 97× speedup while preserving high quality (98% relevance) within the top 40% pushdown.
- 11 February 2026
- Published by Elsevier in Data & Knowledge Engineering
- Vol. 164
When temporary results meet intermediate index: An optimization technique of procedural SQL query processing
- 11 February 2026
- Published by Elsevier in Data & Knowledge Engineering
- Vol. 164
When temporary results meet intermediate index: An optimization technique of procedural SQL query processing
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 128780-128808
Semantic Skyline : Multi-Embedding Skyline Query Processing for RAG Workloads in Vector Databases
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 128780-128808
Open AccessSemantic Skyline : Multi-Embedding Skyline Query Processing for RAG Workloads in Vector Databases