by Adarsh Patil on Nov 11, 2026 |

Call for Presentations:

APEX-AI 2027 @ HiPEAC

January 19, 2027 in Glasgow, United Kingdom
URL: https://apex-ai-workshop.github.io/hipeac27/

Final Submission Deadline
November 15, 2026

Workshop on Approximate, Low-Precision, Sparsity, and Efficient AI Execution from Training to Deployment (APEX-AI 2027)
Co-located with HiPEAC 2027
January 19, 2027 | Glasgow, United Kingdom
https://apex-ai-workshop.github.io/hipeac27/

The growing computational and memory demands of modern AI models call for new approaches to efficient execution across algorithms, compilers, runtimes, and hardware.
APEX-AI brings together researchers and practitioners in computer architecture, systems, machine learning, and design automation to explore approximate, low-precision, sparse, and adaptive techniques for efficient AI training and inference.
The workshop combines invited talks from academia and industry with contributed technical presentations and discussions.

Topics of Interest

  • Low-Precision and Approximate AI: FP8, FP4, INT4 and beyond; mixed and adaptive precision; approximate arithmetic; accuracy–efficiency trade-offs.
  • Sparsity and Conditional Execution: structured/unstructured sparsity, Mixture-of-Experts, dynamic routing, token/layer skipping, and input-dependent computation.
  • Memory and Data Movement: KV-cache optimization, compression, memory-efficient execution, and bandwidth/interconnect efficiency.
  • Efficient Large-Model Execution: LLM inference and training, speculative decoding, long-context processing, distributed and disaggregated inference.
  • Accelerator–Compiler–Runtime Co-Design: domain-specific and heterogeneous accelerators, compiler optimizations, scheduling, and cross-layer hardware–software optimization.

Submission Guidelines
We invite one-page abstracts proposing technical presentations.
Submissions may describe original research, work in progress, preliminary results, previously published research, practical experiences, or emerging research directions.
Submit abstracts: https://easychair.org/conferences/?conf=apexai27

APEX-AI is non-archival. No formal proceedings are planned, and previously published work is explicitly welcome.
Selected contributors will be invited to deliver a 15–20 minute presentation followed by discussion. At least one presenter must register for HiPEAC 2027 and attend in person.

Important Dates

  • Abstract submission: November 15, 2026 (AoE)
  • Notification: December 8, 2026
  • Workshop: January 19, 2027