Compress What You See, Not What You Say: Anchored Context Distillation for Latent-Observation Software Engineering Agents
cs.AI
Submitted: 2026-09-25
Updated: 2026-09-25
Terminology
Sources
- A General Language Assistant as a Laboratory for Alignment
- CoACT: Action-Preserving Observation Compression for Coding Agents
- LatentMem: Customizing Latent Memory for Multi-Agent Systems
- Distilling the Knowledge in a Neural Network
- End-to-End Context Compression at Scale
- The Complexity Trap: Simple Observation Masking Is as Efficient as LLM Summarization for Agent Context Management
- MemGPT: Towards LLMs as Operating Systems
- Qwen3 Technical Report
- A Self-Evolving Framework for Efficient Terminal Agents via Observational Context Compression
- Learning by Distilling Context
- Scaling Long-Horizon LLM Agent via Context-Folding
- SWE-Pruner: Self-Adaptive Context Pruning for Coding Agents
- SWE-Pruner Pro: The Coder LLM Already Knows What to Prune
- ContextWeaver: Selective and Dependency-Structured Memory Construction for LLM Agents
- AgentFold: Long-Horizon Web Agents with Proactive Context Management
- MemGen: Weaving Generative Latent Memory for Self-Evolving Agents
- Mem-W: Latent Memory-Native GUI Agents
- Fine-Tuning Language Models from Human Preferences
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection