Program

Presentations

Oral Presentations

  • RAISE: Loss-Aware, Direction-Free Retrieval over Hierarchical Documents

    Presented by Joohyung Yun

    POSTECH Oct 31, 11:00 – 11:20

  • Pirka: Learning-based Selection of Primitives for Efficient and Robust Subgraph Matching

    Presented by Genryu Kuraya

    UOsaka Oct 31, 11:20 – 11:40

  • CADENZA: Compiling Natural-Language Intent into Task-Specific Operator DAGs for Semantic Query Processing

    Presented by Jaehyun Ha

    POSTECH Oct 31, 11:40 – 12:00

  • Why Benign Post-Training Induces Safety Gate Collapse

    Presented by Zihan Zhao

    UOsaka Nov 1, 09:30 – 09:50

  • AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces

    Presented by Sungho Park

    POSTECH Nov 1, 09:50 – 10:10

  • Language Models Cannot Tell When a Definition Is Missing

    Presented by Zengqing Wu

    UOsaka Nov 1, 10:10 – 10:30

  • Cache Replacement in Direct-Access Cloud OLAP—An Experimental Comparison and Practical Guidance

    Presented by Taejun Yoon

    POSTECH Nov 1, 10:45 – 11:05

  • Random-Walk-Based Cardinality Estimation for Graph Queries with Projection

    Presented by Takaya Miki

    UOsaka Nov 1, 11:05 – 11:25

Poster Presentations

  • Towards Unified Semantic Engine

    Presented by Prof. Kyoungmin Kim

    KAIST

  • CANOPY: Adaptive-Granularity Evidence Compression for Multimodal RAG

    Presented by Hyojeong Yun

    POSTECH

  • Hi-Q: Hierarchical Evidence-guided Query Refinement for Multi-Hop Question Answering

    Presented by Jueun Kim

    POSTECH

  • A Fast Database Replay System with a Fine-Granular Dependency Graph

    Presented by Wonseok Lee

    POSTECH

  • Mnemonica: building reusable research memory upon llm conversation logs

    Presented by Kaina Anderson

    UOsaka

  • Schema Optimization for Time-Dependent Workloads via Materialized View Selection

    Presented by Kanji Masuda

    UOsaka

  • DualBridge: Bridging the Gap in GNN-to-MLP Distillation

    Presented by Daito Ochi

    UOsaka

  • zkMPNN: Query-Local Zero-Knowledge Proofs of Complete Message-Passing Inference

    Presented by He Lyu

    UOsaka

  • MotiveMob: Motivation as Semantic Action for Closed-Loop Human Mobility Generation

    Presented by Mengkun Gao

    UOsaka

  • ToolBottleneckBench: Measuring Tool-Agent Performance under Controlled Pressures and Fixed Harness Budgets

    Presented by Zichuan Xu

    UOsaka

  • How do LLM Bidders Play Black-Box Auctions?

    Presented by Rento Hosokawa

    UOsaka

  • Benchmarking the Dynamics of Popularity Debiasing in Recommender Feedback Loops

    Presented by Yutaka Kojima

    UOsaka