Search papers, labs, and topics across Lattice.
4
0
7
0
Dual Virtual Hubs enable ICFlowNet to achieve a remarkable 9.13-point improvement in long-range F1 scores, revolutionizing how we predict indirect control-flow in stripped binaries.
LLM evaluators use a two-stage mechanism where local error comparisons and rating crystallization happen in distinct layers, reshaping our understanding of their decision-making processes.
LLMs can reveal critical information from OOXML documents that is invisible in Microsoft Office, with up to 76% of trials exposing hidden facts.
Task-aware neuron steering in VLMs is now possible without gradients, unlocking better performance and interpretability across diverse multimodal tasks.
