Trajectory shortcut trees improve agents without outcome labels or annotations

DENSE converts prior agent executions into structured feedback that helps fresh attempts complete terminal tasks more accurately while using fewer recipient-model tokens.

Top University
Siyuan Liu (Fudan University) · Fan Yu (Fudan University) · Dongyu Ru (Meituan Longcat Team) · Yizhu Liu (Meituan Longcat Team) · Yifan Yang (Meituan Longcat Team) · Xuezhi Cao (Meituan Longcat Team) · +2 more
Research Digest··2 min read
Liu et al.

” The method compresses repeated attempts, summarizes successfully completed branches, preserves detail for unresolved branches, and uses evidence of later recovery to reconcile errors identified at different subtask levels.

Why this paper

From Fudan University and Meituan Longcat Team · Part of Agent Self-Improvement, now 10 papers

In one line

DENSE organizes agent trajectories into nested shortcut trees to improve task success without external rewards.

What we could check

  • ·No code link found
  • ·No weights link found
  • ·No dataset link found
  • ·No compute details found
  • ·No stated limitations found
  • ·No benchmark numbers found

Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.

§
newspaper

Research Digest

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.