WritingAmazon (Nova)Amazon (Nova)published Jul 9, 2026seen 2w

Capturing token IDs during agentic interactions for better reinforcement learning

Open original ↗

Captured source

source ↗

Capturing token IDs during agentic interactions for better reinforcement learning - Amazon Science

Close

Close

Social

bluesky

threads

twitter

instagram

youtube

facebook

linkedin

github

rss

Menu

Research

Research areas

Automated reasoning

Cloud and systems

Computer vision

Conversational AI

Economics

Information and knowledge management

Machine learning

Operations research and optimization

Quantum technologies

Robotics

Search and information retrieval

Security, privacy, and abuse prevention

Sustainability

Our scientific contributions

Publications

Research from our scientists and collaborators.

Conferences

Our experts present and discuss cutting-edge research at scientific meetings globally.

Research areas

Automated reasoning

Cloud and systems

Computer vision

Conversational AI

Economics

Information and knowledge management

Machine learning

Operations research and optimization

Quantum technologies

Robotics

Search and information retrieval

Security, privacy, and abuse prevention

Sustainability

Our scientific contributions

Publications

Research from our scientists and collaborators.

Conferences

Our experts present and discuss cutting-edge research at scientific meetings globally.

News & blog

The latest from Amazon researchers

Amazon Science Blog

Technical deep-dives and perspectives from our scientists.

News

Research milestones and recent achievements.

The latest from Amazon researchers

Amazon Science Blog

Technical deep-dives and perspectives from our scientists.

News

Research milestones and recent achievements.

Collaborations

Amazon Research Awards

Overview

Call for proposals

Latest news

Research stories

Recipients

Amazon Nova AI Challenge

Overview

Rules

FAQs

Teams

Research collaborations

Overview

Carnegie Mellon University

Columbia University

Hampton University

Howard University

IIT Bombay

Johns Hopkins University

Max Planck Society

MIT

Tennessee State University

University of California, Los Angeles

University of Illinois Urbana-Champaign

University of Southern California

University of Texas at Austin

Virginia Tech

University of Washington

Amazon Research Awards

Overview

Call for proposals

Latest news

Research stories

Recipients

Amazon Nova AI Challenge

Overview

Rules

FAQs

Teams

Research collaborations

Overview

Carnegie Mellon University

Columbia University

Hampton University

Howard University

IIT Bombay

Johns Hopkins University

Max Planck Society

MIT

Tennessee State University

University of California, Los Angeles

University of Illinois Urbana-Champaign

University of Southern California

University of Texas at Austin

Virginia Tech

University of Washington

Resources

Code and datasets

AGI Labs

Meet the team building useful AI agents.

Amazon Nova

Try Amazon’s frontier foundation models.

Code and datasets

AGI Labs

Meet the team building useful AI agents.

Amazon Nova

Try Amazon’s frontier foundation models.

Careers

Careers

Explore our open roles.

Amazon Scholars

Faculty research opportunities on industry-scale technical challenges.

Postdoctoral Science Program

Early-career research opportunities alongside experienced industry scientists.

Careers

Explore our open roles.

Amazon Scholars

Faculty research opportunities on industry-scale technical challenges.

Postdoctoral Science Program

Early-career research opportunities alongside experienced industry scientists.

Search

Submit Search

Machine learning

Capturing token IDs during agentic interactions for better reinforcement learning

A new Rust proxy called Turnstile sits between the model backend and the agent harness to capture information lost in mere text transcripts.

By Frederick Robinson

July 9, 2026

10 min read

Share

Share

Copy link

Email

X

LinkedIn

Facebook

Line

Reddit

QZone

Sina Weibo

WeChat

WhatsApp

分享到微信

x

Reinforcement learning (RL) is one of the techniques we use to make language models better at sustained, multistep tasks like writing code, navigating a website, or carrying out a research workflow. The model doesn't act alone in those settings; it's wrapped in a piece of software we call a harness , which lets it call tools, observe the results of using them, and decide what to do next. To improve such a model with RL, we let it attempt many tasks inside the harness, score how well each attempt went, and use the score to nudge the model's parameters toward the choices that worked. The hard part turns out to be the bookkeeping. To turn a scored attempt into a parameter update, the trainer needs an exact record of what the model produced — not a summary, and not a transcript that looks as though it captures a complete exchange but drops vital information. Internally, models see text as a sequence of numbered units called tokens ; an English sentence might have 10 or 20 tokens, each assigned an integer ID by a piece of software called a tokenizer. Two strings that look identical in a transcript can map to different token IDs after a small change of formatting, and that gap, however small, is enough to make the trainer optimize against a slightly different past than the one the model actually experienced. Today we're releasing Turnstile , a small proxy written in the Rust programming language that sits between any agent harness and the backend system that runs the model. Turnstile records the exact token-level history of every request as it happens, at the only point where that history is unambiguously correct: the moment of generation. It then exports a generic, framework-neutral trajectory that can feed into whichever RL training stack you already use.

Try Turnstile yourself

Turnstile is open source and available now. Point your existing agent harness at it and start capturing token-native rollout data for RL training.

We use Turnstile to drive real RL training runs. In the validations we report here, two different agents — a text-only coding agent and a multimodal computer-use agent — improved steadily over the course of their RL runs. In both cases the agent harness was left unchanged, and the data Turnstile records flowed directly into the training stack and produced the expected learning signal end to end.

Tokens, rollouts, and why agent transcripts can lie

Three pieces of vocabulary do most of the work in the rest of this post, so it's worth grounding them up front. A tokenizer is a deterministic function that turns text into a list of integer token IDs...

Excerpt shown — open the source for the full document.