The field of electronic design automation (EDA) is moving towards more sustainable and environmentally friendly solutions. Researchers are exploring ways to reduce the carbon footprint of machine learning systems and hardware design. One of the key areas of focus is the development of carbon-aware architectures and design frameworks that can optimize for both performance and environmental impact.

Another significant direction in EDA is the application of representation learning and AI-powered design methods. These approaches have shown promising results in improving the efficiency, accuracy, and scalability of integrated circuit design flows.

Noteworthy papers in this area include:

• Carbon Aware Transformers Through Joint Model-Hardware Optimization, which proposes a framework for sustainability-driven co-optimization of ML models and hardware architectures.
• CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design, which introduces a multimodal and implementation-aware circuit encoder that supports different downstream circuit design tasks.
• QiMeng-CPU-v2: Automated Superscalar Processor Design by Learning Data Dependencies, which presents a novel approach to automatically designing superscalar processors using a hardware-friendly model.
• OneDSE: A Unified Microprocessor Metric Prediction and Design Space Exploration Framework, which offers an efficient and novel design space exploration solution for modern CPU design.
• DiffPattern-Flex: Efficient Layout Pattern Generation via Discrete Diffusion, which generates reliable layout patterns efficiently using a discrete diffusion model.
• Enhancing Reinforcement Learning for the Floorplanning of Analog ICs with Beam Search, which presents a hybrid method that combines reinforcement learning with a beam search strategy for floorplanning of analog ICs.