Decoded
Feed
Paper
Pulse
Search
Listen
Feed
ToDoPro:
Faster
and
Better
Reference-Based
Generation:
Early
Token
Reduction
via
Downsampled
Self-Probing
arXiv:2610.34782
CV
AI
26 oct 2026
1 min read
PDF
Save
Not decoded yet
in the queue
This paper is waiting its turn. The original abstract is at the bottom of the page.
From the paper
Original abstract