uncha-hyperbolic-compositional-alignment

Swap fixed entailment loss with uncertainty-guided piecewise-continuous formulation to improve part-to-whole compositional alignment in hyperbolic vision-language models by +3.0-3.1% on zero-shot classification. Use when training hyperbolic VLMs on part-object relationships and need better multi-object compositional understanding.

Independent GitHub snapshot · Automated classification · Original documentation governs