Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
3
3
5
0
The DeepSeek-V4.1-Flash model, a multimodal Mixture-of-Experts model with 552B backbone parameters and support for contexts of up to one million tokens, is introduced, substantially improving cost efficiency for agentic workloads and pushing the limits of KV cache compression.
Breaking the tight coupling between feature engineering and model architecture, PG enables rapid iterations and deployment in generative retrieval systems, yielding measurable performance gains.
Collision-free document identifiers that boost retrieval performance and business value could redefine how we approach generative retrieval systems.