Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
1
0
2
2
LLMs choke on long numerical sequences, but a simple separator token trick can boost accuracy by 35% and cut token costs by 16%—without any training.