Token Pruning Efficiency
Methods for identifying and removing low-value tokens in language and vision-language models to reduce computational cost while maintaining reasoning performance.
2 papers
Where this stands
The written synthesis of this thread is for subscribers. Subscribe.