Search papers, labs, and topics across Lattice.
Affiliation:
3
0
7
1
Self-correction in LLMs can be dramatically improved by reinforcing step-level reasoning, achieving higher accuracy and reliability in outputs.
TIGER achieves remarkable speedup in multimodal generation by intelligently routing visual tokens based on textual context, outpacing traditional methods.
Uncertainty estimates from LLMs often fail when they're needed most – in low-information scenarios – but a simple post-hoc calibration can fix them.
