# Efficient On-Device Agents via Adaptive Context Management

arXiv: https://arxiv.org/abs/2511.03728 · HTML: https://mlsanity.com/paper/2511.03728/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Efficient reasoning and agent inference, one of 12 directions in Efficient LLM inference and serving.
- Builds on 25 of its 37 references that are in the map.
- First posted on arXiv 24 Sep 2025.

## Placed in

- Efficient LLM inference and serving / [Efficient reasoning and agent inference](https://mlsanity.com/inference/efficient-reasoning-and-agent-inference/)

## Builds on (most cited first, 25 of 25)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs](https://mlsanity.com/paper/2307.16789/) (2023)
- [LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://mlsanity.com/paper/2403.13372/) (2024)
- [OpenHands: An Open Platform for AI Software Developers as Generalist Agents](https://mlsanity.com/paper/2407.16741/) (2024)
- [ToolAlpaca: Generalized Tool Learning for Language Models with 3000 Simulated Cases](https://mlsanity.com/paper/2306.05301/) (2023)
- [MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents](https://mlsanity.com/paper/2506.15841/) (2025)
- [APIGen-MT: Agentic Pipeline for Multi-Turn Data Generation via Simulated Agent-Human Interplay](https://mlsanity.com/paper/2504.03601/) (2025)
- [XAttention: Block Sparse Attention with Antidiagonal Scoring](https://mlsanity.com/paper/2503.16428/) (2025)
- [SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks](https://mlsanity.com/paper/2503.15478/) (2025)
- [xLAM: A Family of Large Action Models to Empower AI Agent Systems](https://mlsanity.com/paper/2409.03215/) (2024)
- [Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning](https://mlsanity.com/paper/2410.19258/) (2024)
- [KVzip: Query-Agnostic KV Cache Compression with Context Reconstruction](https://mlsanity.com/paper/2505.23416/) (2025)
- [Learning to Break the Loop: Analyzing and Mitigating Repetitions for Neural Text Generation](https://mlsanity.com/paper/2206.02369/) (2022)
- [Recursively Summarizing Enables Long-Term Dialogue Memory in Large Language Models](https://mlsanity.com/paper/2308.15022/) (2023)
- [TinyAgent: Function Calling at the Edge](https://mlsanity.com/paper/2409.00608/) (2024)
- [Hammer: Robust Function-Calling for On-Device Language Models via Function Masking](https://mlsanity.com/paper/2410.04587/) (2024)
- [Granite-Function Calling Model: Introducing Function Calling Abilities via Multi-task Learning of Granular Tasks](https://mlsanity.com/paper/2407.00121/) (2024)
- [KeyDiff: Key Similarity-Based KV Cache Eviction for Long-Context LLM Inference in Resource-Constrained Environments](https://mlsanity.com/paper/2504.15364/) (2025)
- [HAICOSYSTEM: An Ecosystem for Sandboxing Safety Risks in Human-AI Interactions](https://mlsanity.com/paper/2409.16427/) (2024)
- [Octopus: On-device language model for function calling of software APIs](https://mlsanity.com/paper/2404.01549/) (2024)
- [Collaborating with AI Agents: Field Experiments on Teamwork, Productivity, and Performance](https://mlsanity.com/paper/2503.18238/) (2025)
- [Arch-Router: Aligning LLM Routing with Human Preferences](https://mlsanity.com/paper/2506.16655/) (2025)
- [On-Device LLM for Context-Aware Wi-Fi Roaming](https://mlsanity.com/paper/2505.04174/) (2025)

## Built on by (most cited first, 1 of 1)

- [Actionable Hallucination Detection: Translating Latent Uncertainty into Agentic Critique](https://mlsanity.com/paper/2608.10430/) (2026)

## Abstract (arXiv)

On-device AI agents offer the potential for personalized, low-latency assistance, but their deployment is fundamentally constrained by limited memory capacity. Context in agentic settings worsens this problem due to large static tool schemas and a growing interaction history that continually expands the persistent KV cache. To maintain on-device feasibility, agents must operate near the minimum task-sufficient context, while preserving task performance. We introduce two complementary mechanisms: (1) a learned intra-session memory architecture that distills trajectories into an append-only Context State Object (CSO), preserving current state and relevant details from previous steps that may have future utility, while still supporting KV-cache reuse, and (2) a just-in-time schema-passing mechanism that loads full tool definitions only upon tool selection. We instantiate this framework by adapting 3B-parameter SLMs and profiling them on smartphone hardware. In our evaluations, CSO-based memory outperforms KV-cache compression and trained summarization while retaining performance near or even above full-history execution. Deployed on-device, our framework reduces initial tool context by over 6× and interaction-context growth by 10-25 times, substantially reducing initial