Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
Affiliation:
5
0
6
Embedding Lorentz transformations directly into query-key attention cuts video prediction error by nearly 10% using a tenth of the parameters, proving that explicit relativistic spacetime geometry can dramatically outperform purely learned positional encodings across physical scales.
AMM designers face a stark choice: build a market that's fair, efficient, or incentivizes good behavior, because you can only pick two.
MiniMax-M2 proves that massive parameter counts don't always translate to better agentic performance; strategic activation of a smaller subset can unlock frontier-level intelligence.
Transferring policies from coarse, incomplete simulators to the real world is possible by correcting the simulator's dynamics with real-world data.