Open Models Almanac
Methodstraining objectivedistillation objective

single mention · training objective

Distillation-loss alignment of VLM queries with depth tokens

VLM learnable queries are aligned with LingBot-Depth tokens by minimizing a distillation loss.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See distillation objective for the methods filed beside it.

Filed under training objective › distillation objective