Search papers, labs, and topics across Lattice.
3
0
7
Late-interaction models can outperform dense models in multilingual retrieval, revealing a new avenue for enhancing cross-lingual performance.
Turns out, you can boost a multilingual LLM's reasoning in any language just by swapping in the mid-layers from an English-reasoning specialist.
Forget knowledge distillation: pre-training ColBERT from scratch on public data alone beats models distilled from stronger, closed-source single-vector baselines.
