Tag
This paper introduces retrieved-span training for efficient query-focused meeting summarization on the QMSum dataset, showing that fine-tuning on retrieved spans recovers performance loss and that a smaller 406M model can achieve comparable results to larger models using automatic metrics.
InclusionAI has open-sourced the Ling-3.0 series, featuring highly efficient language models with sparse MoE architecture and hybrid linear attention, providing checkpoints at various training stages to support research and innovation.
Introduces Efficient Operator Search (EOS), a unified differentiable framework that generalizes token reduction methods (pruning, merging, pooling, adaptive reweighting) into a shared operator space, automatically searching for optimal operator compositions under budget constraints. The method achieves competitive results across benchmarks and reveals consistent operator patterns.
Introduces FaraGen, a synthetic data generation system for computer use agents, and Fara-7B, a small but efficient model that outperforms larger counterparts on web task benchmarks. The model is released open-weight on Microsoft Foundry and HuggingFace.