Skip to content
NewA calmer, faster ZeroHour

CarryOn organizes long-horizon agent context around task objects

Original title: Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents

Hugging Face daily papersby Yuchen Lu, Julian McAuley, Lei Zhang ()1 min read1 · 1 read
AI summary · x-ai/grok-4.7Summaries by Vani AI

Authors report gains of 5.8% and 7.0% with token cuts of 51.9% and 22.9% on two benchmarks.

CarryOn is an object-oriented context runtime that continues long-horizon agents from reusable task objects rather than full interaction histories. It keeps continuation carriers, complementary execution context, and a live tail of recent interactions, updating only affected carrier records after history is retired. On OSWorld 2.0 and BrowseComp-Plus, the authors report gains of 5.8% and 7.0% while cutting total tokens by 51.9% and 22.9%.

Unconfirmed: Benchmark gains and token reductions are author-reported and not independently confirmed.

Key facts
  • Context is built from carriers, execution context, and a recent-interaction tail.
  • Unaffected carrier records are carried forward across folds.
  • OSWorld 2.0: 5.8% gain and 51.9% fewer tokens, per the authors.
  • BrowseComp-Plus: 7.0% gain and 22.9% fewer tokens, per the authors.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

1 min read at huggingface.co

Long-horizon agents accumulate progress in persistent task state, yet repeatedly process interaction histories describing how that progress was achieved. This raises a natural question: can long-horizon context instead be organized around reusable progress rather than the history that produced it?

Hugging Face daily papers

By Yuchen Lu, Julian McAuley, Lei Zhang · huggingface.co

Read the full article at Hugging Face daily papers

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.