Track synthesis, materials and catalysis as it is published.
Search papers, labs, and topics across Lattice.
Carnegie Mellon University, NVIDIA & Massachusetts Institute of Technology
CMU Machine Learning2
0
4
Native 8-bit LLM training silently destabilizes at billion-parameter scale due to softmax gradient drift, but restoring the zero-row-sum invariant completely eliminates the FP8 attention gap without architecture modifications.
Jet-Long achieves up to 1.39x throughput improvements while maintaining accuracy across long-context tasks, setting a new standard for zero-shot context extension in LLMs.