Priyadarshini Panda
University of Southern California
Algorithm-hardware co-design and neuromorphic computing for ultra-low-power AI systems.
Confronting the computational costs of scaling Language, Vision, and Action Models.
SELVA focuses on efficiency and sustainability across language, vision, and action models, especially for real-time and resource-constrained deployment.
Recent advances in LLMs, VLMs, and VLA systems are accompanied by rapidly increasing compute and energy costs. SELVA aims to foster cross-disciplinary discussion spanning NLP, computer vision, robotics, and systems to develop methods that are both high-performing and computationally feasible.
We welcome high-quality original research and system papers related to efficient and sustainable LVA models.
Hybrid format planned: invited talks, contributed talks, poster session, and award ceremony.
All submissions must be in PDF format, anonymized, and prepared using the ACL Article Template. This workshop follows the ACL formatting guidelines: long papers may use either the Research Track or Industry Track template, while abstracts and short papers should follow the same format with a limit of 4 pages (including figures, tables and references). Authors should also follow the ACL 2026 author instructions .
May 25, 2026
Jun 10, 2026
Jun 15, 2026
July 03, 2026
Timezone convention: AoE (UTC−12).
Tentative invited speakers for the SELVA Workshop.
University of Southern California
Algorithm-hardware co-design and neuromorphic computing for ultra-low-power AI systems.
UC Berkeley
Practical challenges in training and deploying efficient deep neural networks.
ETH Zürich
TinyML, smart sensing, and energy-efficient IoT systems for low-power edge intelligence.
NVIDIA
LongLoRA / LongVILA and efficient long-context modeling for interactive agents.
University of Arizona
Mathematical foundations of efficient deployment and adaptation for robust AI models.
Beihang University / SenseTime Research
Model quantization, sparsity, and deployment systems for efficient large-model inference.
Core organizers and contributors listed in the submission.
Initial PC members listed in the submission excerpt (can be extended on final website).
ACL 2026 is scheduled in San Diego, and workshop logistics can align with the ACL workshop venue details.
The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026)
San Diego, California, USA · July 2–7, 2026
Workshop days: July 3–4, 2026 (SELVA exact date/room TBD).
Grand Hyatt Manchester San Diego
1 Market Place, San Diego, CA 92101