LLM Agents Are Latent Context Managers: Eliciting Self-Managed Context via State Proprioception
Binyan Xu, Haitao Li, Kehuan Zhang
Why It Matters
What makes this one worth your time
This work addresses a critical limitation in long-horizon tool agents, potentially improving their effectiveness in real-world applications where context management is crucial.
VISTA enhances context management in LLMs by making internal state visible and manageable.
Summary
The paper introduces VISTA, a model-agnostic interface that enhances context management in language models by providing visibility into internal state, thereby improving performance on various benchmarks without requiring training.
Key contributions
- Introduction of VISTA, a training-free interface for context management in language models.
- Demonstration of significant performance improvements on multiple benchmarks with the same untrained interface.
- Provision of a runtime dashboard that visualizes token usage, recency, and access history for better context management.
Notable insights
- The concept of 'state proprioception' highlights a previously unaddressed aspect of context management in language models.
- The model-agnostic nature of VISTA suggests broad applicability across different architectures without the need for retraining.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2606.30005v4 Announce Type: replace Abstract: Long-horizon tool agents are bottlenecked by how their context grows toward the limits of the context window. Recent systems make context management agent- or system-controlled, but they either learn a compression policy that discards evidence or manage context in a layer the agent never sees. We argue both leave a more basic gap unaddressed. Frontier language models are proprioceptively blind to their own context. From the prompt alone they cannot see how large, how old, or how used each block is, the signals a keep-or-drop decision needs. We hypothesize that competent context management is already latent in capable models, and that what is missing is not a learned policy but an interface exposing this state. We introduce VISTA (Visible Internal State for Tool Agents), a training-free, model-agnostic layer that represents working memory as typed, addressable blocks, surfaces a runtime dashboard of per-block token usage, recency, and access history, and archives blocks as recoverable full-fidelity payloads. On LOCA-Bench, BrowseComp-Plus, and GAIA, the same untrained interface transfers across 1M-, 100K-, and 10K-scale trajectories. On LOCA-Bench it improves four backbones and lifts Gemini-3-Flash from 22.7 to 50.7%. The lift grows with context pressure and transfers across backbones. Ablations further confirm that the dashboard matters beyond archive and recovery tools.