Marlin Sustainable Llm Inference

---
name: marlin-sustainable-llm-inference
description: 'Research paper: MARLIN: Multi-Agent Game-Theoretic Reinforcement Learning for Sustainable LLM Inference in Cloud Datacenters.'
metadata:
  openclaw:
    emoji: "⚡"
    tags: ["research", "arxiv", "multi-agent-rl", "llm", "sustainability", "game-theory"]
---

# MARLIN: Multi-Agent Game-Theoretic Reinforcement Learning for Sustainable LLM Inference in Cloud Datacenters

**arXiv ID:** 2605.13496
**Published:** 2026-05-13
**Authors:** H. Moore, S. Qi, D. Milojicic, C. Bash, S. Pasricha
**Categories:** cs.DC, cs.LG
**Utility Score:** 0.95

## Abstract

Large Language Models (LLMs) have become increasingly prevalent in cloud-based platforms, propelled by the introduction of AI-based consumer and enterprise services.
LLM inference requests in particular account for up to 90% of total LLM lifecycle energy use, dwarfing training energy costs. The rising volume of LLM inference requests is increasing environmental footprints, particularly carbon emissions and water consumption.
To improve sustainability for LLM inference serving in cloud datacenter environments, we propose a novel multi-agent game-theoretic reinforcement learning framework called MARLIN to co-optimize time-to-first token (TTFT), carbon emissions, water usage, and energy costs associated with LLM inference.
MARLIN demonstrates a reduction of at least 18% in TTFT, 33% in carbon emissions, 43% in water usage, and 11% in energy costs compared to state-of-the-art LLM inference management frameworks.

## Key Contributions

1. **MARLIN Framework**: Proposed a novel multi-agent game-theoretic reinforcement learning framework (MARLIN) for sustainable LLM inference in cloud datacenters
2. **Multi-objective Optimization**: MARLIN co-optimizes time-to-first token (TTFT), carbon emissions, water usage, and energy costs associated with LLM inference
3. **Significant Improvements**: Demonstrates reductions of at least 18% in TTFT, 33% in carbon emissions, 43% in water usage, and 11% in energy costs compared to state-of-the-art LLM inference management frameworks

## Relevance to AI Systems

- **Sustainable AI**: Addresses the growing environmental impact of LLM inference, which accounts for up to 90% of total LLM lifecycle energy use
- **Game-Theoretic MARL**: Introduces game-theoretic approaches to multi-agent coordination in LLM systems for optimizing conflicting objectives
- **Cloud AI Systems**: Provides a framework for optimizing LLM serving in cloud datacenter environments with consideration for computational efficiency and environmental sustainability

## Technical Keywords

multi-agent reinforcement learning, game theory, LLM inference, sustainability, cloud computing, TTFT, carbon emissions, water usage, energy costs

## URL

https://arxiv.org/abs/2605.13496

---

**Tracked:** 2026-05-30
**Source:** arXiv Paper Tracker