Search papers, labs, and topics across Lattice.
2
0
4
3
A unified model that seamlessly integrates brain, vision, and language reveals new insights into multimodal neural processing and outperforms traditional methods.
Test-time training can finally scale for large reasoning models: TEMPO unlocks sustained performance gains by interleaving policy refinement with periodic critic recalibration, boosting accuracy by over 18% on challenging benchmarks.
