Search papers, labs, and topics across Lattice.
1
1
0
2
It is suggested that safety does not automatically scale with reasoning capability, especially under long-context inference, as context length increases and safety alignment consistently degrades.
