Search papers, labs, and topics across Lattice.
Southwest Minzu University
1
0
3
2
Forget finetuning – structured self-reflection alone can boost VLM reasoning robustness, surpassing even larger models on OCRBench v2.
