Another research accepted to ICLR 2026 We explored a new way to shrink long prompts using smaller draft models from different model families, no retraining needed. Faster time to first token, with performance holding strong. Take a look @UrmishThakker

→ View original post on X — @sambanovaai,