Brief

Researchers have developed DOPPLER, a novel three-stage framework for optimizing device assignment in asynchronous dataflow graphs, particularly for complex machine learning workloads. This system addresses limitations of previous methods by supporting asynchronous systems and integrating both reinforcement learning and expert-designed heuristics. DOPPLER's dual-policy network, comprising selection and placement policies, has demonstrated superior performance in reducing execution time and improving training efficiency compared to existing baselines. AI

Researchers have developed MHA-RAG, a novel framework that encodes domain-specific examples as soft prompts rather than traditional text. This approach, utilizing Multi-Head Attention, aims to improve the efficiency and accuracy of adapting foundation models to new domains with limited data. Experiments show MHA-RAG achieves a 20-point performance gain over standard RAG while reducing inference costs by 10x, demonstrating superior accuracy and efficiency regardless of exemplar order. AI