Databases

Yanlin Qi, Xinhang Chen, Huiqiang Jiang et al. · harvard, microsoft-research

This work provides a significant improvement in the efficiency and scalability of long-context LLM inference for developers and researchers working with large language models.

Jiongli Zhu, Haoquan Guan, Parjanya Prajakta Prashant et al.

For practitioners of Text-to-SQL, this provides a method to build more reliable ensembles without model fine-tuning, addressing the bottleneck of correlated failures in candidate generation.

Yuxuan Zhu, Tengjun Jin, Yoojin Choi et al.

This work addresses the critical challenge of improving Text-to-SQL accuracy for database research and data analytics applications, representing a significant advancement rather than an incremental improvement.

Rongsheng Zhang, Ruofan Hu, Weijie Chen et al.

For developers of role-playing agents, this work addresses the bottleneck of maintaining persona consistency in long conversations, offering a novel memory architecture and benchmark.

Duyi Pan, Tianao Lou, Xin Li et al.

This addresses recall and precision limitations in graph-based RAG for black-box knowledge graphs, offering a plug-and-play solution for knowledge-intensive tasks.

Yixi Zhou, Fan Zhang, Zhiqiao Guo et al.

This addresses the overlooked dimension of structural evaluation for LLM-based program generation systems, which is incremental but important for reliability.

Yougen Zhou, Qin Chen, Ningning Zhou et al.

This work addresses the need for better assessment of reward models in mental health applications, though it is incremental as it focuses on benchmarking rather than developing new methods.

Zihao Zheng, Zhihao Mao, Sicheng Tian et al.

This work addresses inference efficiency for robot control systems using VLA models, offering a hybrid approach that is incremental but provides concrete speed improvements.

Muhammed Saeed, Simon Razniewski

This work addresses the problem of overestimating LLM factuality in benchmarks for researchers and practitioners, highlighting significant gaps in knowledge coverage and accuracy.

Giuseppe Samo, Paola Merlo

This work addresses a gap in evaluating systematic cross-sentence knowledge in LLMs for linguists and NLP researchers, but it is incremental as it focuses on dataset creation and baseline testing without major methodological breakthroughs.

Yanchen Wu, Tenghui Lin, Yingli Zhou et al.

This work addresses the need for standardized evaluation in memory methods for LLM agents, offering incremental improvements through a novel hybrid method.

Jinhu Qi, Yifan Li, Minghao Zhao et al.

This addresses the need for more realistic and representative trustworthiness evaluation for agentic AI systems, which is crucial for safe deployment, though it appears incremental as it builds on existing evaluation concepts.

Zirui Tang, Xuanhe Zhou, Yumou Liu et al.

This benchmark addresses the underexplored problem of evaluating AI agents on real-world workspace tasks with complex file dependencies, revealing that current agents are far from reliable.

Linjie Xu, Yanlin Zhang, Quan Gan et al.

This work enables zero-shot predictive modeling over relational databases for practitioners, eliminating the need for task-specific model training.

Mingwei Ye, Jiaxi Zhuang, Mingjun Xu et al.

This work solves the problem of democratizing access to document-centric databases for users by enabling more accurate and scalable natural language to query conversion, though it appears incremental as it builds on existing refinement and feedback methods.

Mohammadtaher Safarzadeh, Hitesh Laxmichand Patel, Afshin Orojlooyjadid et al.

For researchers evaluating LLMs on NL2SQL tasks, SPENCE provides a method to assess benchmark contamination, ensuring more trustworthy performance comparisons.

Tianshu Zhang, Kun Qian, Siddhartha Sahai et al.

This addresses the robustness issue for text-to-SQL systems in real-world dynamic database environments, though it is incremental as it builds on existing text-to-SQL research with a new benchmark.

Alexander W. Lee, Benjamin Han, Shayak Sen et al.

For users of semantic query engines, Evergreen provides efficient and accurate verification of natural language aggregates, addressing the bottleneck of grounding claims in large relations.

Ryoto Miyamoto, Xin Fan, Hayato Yamana

For data providers deploying Text-to-SQL across trust boundaries, this work provides a deterministic enforcement mechanism for column-use policies, addressing a practical security and compliance need.

Mengru Wang, Haozhe Luo, Zhenqian Xu et al.

For researchers and developers of LLM memory systems, this paper highlights a previously overlooked failure mode and provides a benchmark and mitigation method, though the impact is limited to the memory-augmented LLM domain.