Tag
This survey examines in-context reinforcement learning (ICRL) under non-stationary environments, where a pretrained decision model adapts through accumulated context without parameter updates. It organizes the literature around what changes, how it unfolds, and how observable it is, and identifies research gaps such as stale-context stress tests and adaptive forgetting.
ACE (Agentic Context Engineering) introduces a framework that treats contexts as evolving playbooks, preventing context collapse and improving performance on agent and domain-specific benchmarks. The work highlights the potential of harness engineering as a data engine for model training.
The paper proposes GTBP, a graph-based back-propagation framework for context adaptation in multi-LLM agentic systems, which improves prompt optimization with theoretical convergence guarantees and outperforms existing methods on benchmarks.