Search papers, labs, and topics across Lattice.
3
0
6
This work revisits RAG compression from a data-mining perspective by aggregating historical query--document--model interactions into reusable evidence views and proposes Reusable Evidence View Aggregation (REVA), a framework that mines the target generator's historical attention traces into a document-keyed, budget-agnostic score store.
Enforcing global invariance in domain generalization can lead to the loss of critical predictive factors, but subset-shared invariance allows for more nuanced representation learning that boosts performance.
ChemBack reveals that even with rigorous admission protocols, chemically valid backdoors can still pose significant risks to molecular graph neural networks.
