ALAS: Transactional and Dynamic Multi-Agent LLM Planning

TL;DR

ALAS introduces a novel framework for multi-agent LLM planning that improves reliability and efficiency by separating validation, recording execution logs, and performing localized repairs, outperforming existing methods on classical benchmarks.

Contribution

The paper presents ALAS, a stateful, disruption-aware multi-agent planning framework that enhances robustness and efficiency through independent validation, versioned logs, and minimal localized repairs.

Findings

Achieves 83.7% success rate on benchmarks

Reduces token usage by 60%

Runs 1.82 times faster than comparable baselines

Abstract

Large language models enable flexible multi-agent planning but remain fragile in practice: verification is often circular, state changes are not tracked for repair, and small faults trigger costly global recomputation. We present ALAS, a stateful, disruption-aware framework that separates planning from non-circular validation, records a versioned execution log for grounded checks and restore points, and performs localized repair that preserves work in progress. The validator operates independently of the planning LLM with fresh, bounded context, avoiding self-check loops and mid-context attrition. The repair protocol edits only the minimal affected region under explicit policies (retry, catch, timeout, backoff, idempotency keys, compensation, loop guards) defined in a canonical workflow IR that maps to Amazon States Language and Argo Workflows. On job-shop scheduling suites (DMU, TA) across five classical benchmarks, ALAS matches or exceeds strong single-LLM and multi-agent baselines, achieving 83.7% success, reducing token usage by 60%, and running 1.82times faster under comparable settings. A minimal reliability study shows that the validator detects injected structural faults with low overhead, and that localized repair contains runtime perturbations with a bounded edit radius and less makespan degradation than global recompute. Results indicate that the combination of validator isolation, versioned execution logs, and localized repair provides measurable efficiency, feasibility, and scalability for multi-agent LLM planning. Code and seeds will be released.