Search papers, labs, and topics across Lattice.
1
0
3
4
Forget finetuning: TTA-Vid adapts video reasoning models to new datasets *during inference* using test-time reinforcement learning, achieving state-of-the-art results without any labels.
