The workshop is a half-day event on August 17th combining an invited keynote, contributed paper presentations, and an interactive discussion round. Exact clock times will be confirmed once the IJCAI schedule is finalized. The coffee break slot is set by the conference and may shift.

Note: This program is preliminary. The overall schedule and individual presentation lengths may be subject to changes once IJCAI announces the final session timings.

TimeSession
5 minWelcome & Introduction
35 minKeynote
Prof. Dr. Norbert Wehn 🌐 — RPTU University, Germany
Title TBD
55 minPaper Session I — Power Usage
Three full-paper presentations (15 min talk + 3 min Q&A each)
  1. Safety-Constrained Contextual Bandit for Dynamic Power Management
  2. WattGPU: Predicting Inference Power and Latency on Unseen GPUs and LLMs
  3. WattLayer: Get Layers Right to Estimate Inference Energy of Neural Networks
10 minPosition Paper & Discussion
Short pitch (3 min) followed by an open audience discussion (7 min)
30 min☕ Coffee Break
55 minPaper Session II — Efficient Models & Deployment
Three full-paper presentations (15 min talk + 3 min Q&A each)
  1. CompressKV: Semantic-Retrieval-Guided KV-Cache Compression for Resource-Efficient Long-Context LLM Inference
  2. TinyHybrid: Ultra-Efficient CNN-Transformer Architecture for Edge-Deployable Brain Tumor Classification
  3. ENAS: An Efficient Hardware-Aware Neural Architecture Search Framework for TinyML on Resource-Constrained Microcontrollers
25 minShort Paper Session
Two short-paper presentations (10 min talk + 3 min Q&A each)
  1. Efficient Vision Models for Jetson: Steel Classification via Knowledge Distillation
  2. Accounting for Bias Enables Sustainable LLM Evaluation
5 minClosing Remarks
including the Best Paper Award 🏆