Speculative Actions: A Lossless Framework for Faster AI Agents

ICLROral2026

Authors
Naimeng Ye, Arnav Ahuja, Georgios Liargkovas, Yunan Lu, Kostis Kaffes, Tianyi Peng
Affiliation
Columbia University
Venue
ICLR 2026
Track
Oral

TL;DR

We introduce speculative actions—a lossless framework that predicts likely actions using faster models, enabling multiple API calls to be executed in parallel and thus yields substantial acceleration.

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

agent

← All ICLR 2026 Oral papers · Browse the whole archive