Recoverable long-horizon AI agents — a framework-agnostic reference harness + recovery-faithful live benchmark. Thesis: "Checkpoints Are Compactions" via Re-grounding Recovery. 0.x: v1.0 held until a powered live-LLM study confirms the claims.
benchmark fault-tolerance recovery checkpoint-restart resilience autonomous-agents ai-agents ai-research checkpointing llm durable-execution llm-agents context-management agentic-ai agent-memory code-agent agent-runtime long-horizon-agents agent-recovery re-grounding
-
Updated
Jun 23, 2026 - Python