Open Models Almanac
Methodsinference & servingcontext management

implementation detail · filed under inference & serving

Trajectory summarization and rollout re-initiation

Summary is a context management strategy that summarizes the overflowed trajectory and re-initiates the rollout.1,2 It addresses maximum length limitations in agentic workflows by extending token budgets at test time.3

Trade-off Summary improves performance by extending average steps but has relatively low overall efficiency.4

How this definition was written

Summarised by glm-5.3-flash from 2 sources; every sentence is checked against the quote it cites.

  1. Summary, which summarizes the overflowed trajectory and re-initiates the rollout — paper
  2. Summary, which summarizes the overflowed trajectory and re-initiates the rollout — DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
  3. to extend token budgets at test time — DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
  4. However, its overall efficiency is relatively low — paper

Also called Summary.

sources
2
model
1
labs adopt it
0
strongest
evaluated

Documented in

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
evaluated 2
↑