Accelerating PayPal’s Commerce Agent with Speculative Decoding:

An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models · 2026

Read Accelerating PayPal’s Commerce Agent with Speculative Decoding: by An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-19767.txt.

Description

Qin, Wan, Mudunuri, and Manoharan benchmark EAGLE3 speculative decoding for PayPal's Commerce Agent, which runs on a fine-tuned llama3.1-nemotron-nano-8B-v1 model. Tests run through vLLM against NVIDIA NIM on identical 2xH100 hardware across 40 configurations. The gamma-3 setting lifts throughput by 22 to 49 percent and cuts latency by 18 to 33 percent at no added hardware cost. Acceptance rates hold steady near 35.5 percent for gamma-3 across all conditions, while gamma-5 gives diminishing returns.

Catalog record

CREATORAn Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE24140
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.19767

Browse the full library · About the Tower of Babel