Search papers, labs, and topics across Lattice.
3
0
5
Thinking Collapse can severely impair reasoning in LLMs, but a new adaptive framework boosts accuracy by over 4% while preserving cognitive capacity.
Unlock practical memory savings and faster decoding in LLMs with PrunePath, a structured sparsification method that adaptively activates experts based on token-level probability budgets.
Achieve robust imbalanced classification with scarce minority samples by turning a generative VAE into a discriminative classifier using distribution-aware fine-tuning and statistically sound hypothesis testing.
