Back to feed
arXiv cs.LG·

Graph-based Target Back-Propagation for Context Adaptation in Multi-LLM Agentic Systems

Signal
78
Hype
15
In three linesGTBP (Graph-based Target Back-Propagation) is a context adaptation framework for multi-LLM agentic systems. It back-propagates local targets through a directed acyclic graph workflow and updates prompts stage-wise. Theoretically convergent, outperforms baselines across 3 benchmarks.
Read source
Your take?
AI AgentsMulti-agentPrompt engineeringReasoning

Summary generated by Claude — human-verified