Arbor: Multi-Agent Framework Uses Tree Search to Optimize LLM Inference Performance
Researchers introduced Arbor, a multi-agent framework that uses structured tree search as a coordination mechanism for autonomous agents optimizing LLM inference across hardware and software stacks. The system pairs an Orchestrator agent that delegates tasks with a Critic agent that validates results, enabling fully autonomous multi-day optimization campaigns. Arbor achieved up to 193% inference throughput-latency improvements over vendor baselines, significantly outperforming single-agent approaches and demonstrating reproducibility across hardware generations.
Arbor is a novel multi-agent framework that introduces tree search as a cognition layer for autonomous agents operating in complex, stateful action spaces. Unlike prior optimization systems that work on isolated targets with stateless evaluation, Arbor maintains an explicit search tree of scored hypotheses as shared working memory that evolves with each measurement. The framework decomposes agent capabilities into hard skills (domain expertise) and soft skills (coordination protocols), enabling coordinated optimization across the full LLM inference stack—from application layer through framework, compiler, kernel, and hardware. The system employs a checks-and-balances architecture where an Orchestrator agent drives optimization by delegating to Domain Specialists, while a Critic agent safeguards stability through root-cause analysis and measurement validation. Experimental validation shows Arbor achieves up to 193% Pareto improvement in inference throughput-latency over vendor-optimized baselines, with run-to-run variance within 2 percentage points, demonstrating hardware-agnostic reproducibility.
What's missing
The paper does not discuss computational overhead of running the multi-agent framework itself, comparison to other multi-agent optimization approaches, or limitations in domains outside LLM inference optimization. The study also does not address how the framework scales to even larger action spaces or provide analysis of failure modes beyond the single-agent crash mentioned.
What different sources said
- arXiv cs.AICenter
Arbor: Tree Search as a Cognition Layer for Autonomous Agents
Related
Topology-Aware Thermodynamics Improves DNA Probe Specificity Design
Researchers developed a new framework for designing DNA probes that accounts for the spatial organization of matched sequences, not just overall thermodynamic stability. Traditional methods rely on scalar measures like melting temperature and free energy, which miss how mismatches are distributed along the probe. The approach could improve diagnostic accuracy in applications like HPV detection and gene expression profiling.
Study Identifies Optimal Thermal Dose for Combining Focused Ultrasound with Immunotherapy in Tumors
Researchers used multimodal PET imaging to identify an optimal thermal dose range for focused ultrasound ablation that destroys tumor tissue while preserving conditions for immunotherapy delivery. The study found that excessive heating collapses blood vessels needed for antibody access, while insufficient heating fails to adequately reduce tumor burden. The findings could guide clinical design of combination treatments pairing thermal ablation with immunotherapies.
Plant MSH1 Protein Functions as Mismatch-Directed Nuclease for Organelle Genome Maintenance
Researchers have identified the precise mechanism by which the AtMSH1 protein in Arabidopsis plants recognizes and cleaves DNA mismatches and lesions, preventing mutations in organellar genomes. The protein combines a DNA mismatch recognition module with a nuclease domain that makes staggered cuts at specific positions relative to DNA damage. This discovery explains how plants maintain unusually low mutation rates in their mitochondrial and chloroplast DNA compared to other eukaryotes.