Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
1
0
3
Most AI models miss subtle Chinese character rendering errors, but a new verifier using Ideographic Description Sequences spots them with 84% F1, outperforming OCR and VLMs by a wide margin.