{
  "schema_version": "0.1.0",
  "record_type": "research-topic",
  "topic_id": "llm-efficiency",
  "label": "LLM Efficiency",
  "title": "LLM and Transformer Efficiency Research",
  "description": "Research papers on efficient Transformer and large-language-model training, inference, and deployment.",
  "introduction": "This topic covers methods for reducing the computation, memory, energy, or deployment cost of Transformer and large-language-model workloads. The records identify the efficiency intervention, model or workload, quality metric, resource measure, and trade-off reported by the source.",
  "aliases": [
    "large-language-model efficiency",
    "Transformer efficiency",
    "efficient AI systems"
  ],
  "search_terms": [
    "LLM efficiency",
    "Transformer optimization",
    "model compression",
    "quantization",
    "sustainable AI"
  ],
  "related_topics": [
    "model-compression",
    "deep-learning-systems",
    "energy-efficiency",
    "machine-learning",
    "performance-optimization"
  ],
  "canonical_url": "https://threadslab.org/research-publications/topics/llm-efficiency.html",
  "paper_count": 2,
  "papers": [
    {
      "paper_id": "energy-consumption-analysis-large-language-models-cpu-gpu",
      "title": "Energy Consumption Analysis of Large Language Models Across CPU and GPU Using Diverse Metric Types",
      "year": 2025,
      "authors": [
        "Tong Zhang",
        "Leila Tahmooresnejad",
        "Naser Ezzati-Jivan"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/energy-consumption-analysis-large-language-models-cpu-gpu/",
      "canonical_source_url": "https://doi.org/10.1109/CASCON66301.2025.00056",
      "core_contribution": "The paper models LLM inference energy separately on CPU and GPU using hardware counters, device metrics, and task/model features, then compares classical and neural regressors across language tasks.",
      "tags": [
        "energy-efficiency",
        "llm-efficiency",
        "performance-modeling",
        "machine-learning",
        "performance-analysis"
      ],
      "keywords": [
        "LLM energy consumption",
        "CPU energy",
        "GPU energy",
        "green AI",
        "static metrics",
        "dynamic metrics",
        "NLP workloads",
        "energy prediction",
        "CASCON 2025"
      ]
    },
    {
      "paper_id": "optimization-transformers-llms",
      "title": "Optimization Strategies for Enhancing Resource Efficiency in Transformers & Large Language Models",
      "year": 2025,
      "authors": [
        "Tom Wallace",
        "Beatrice M. Ombuki-Berman",
        "Naser Ezzati-Jivan"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/optimization-transformers-llms/",
      "canonical_source_url": "https://doi.org/10.1145/3676151.3719379",
      "core_contribution": "The paper compares compression and optimization strategies for reducing the resource cost of Transformer and large-language-model workloads while retaining useful accuracy.",
      "tags": [
        "llm-efficiency",
        "energy-efficiency",
        "model-compression",
        "performance-engineering"
      ],
      "keywords": [
        "transformers",
        "quantization",
        "knowledge distillation",
        "pruning",
        "4-bit quantization",
        "Minitron",
        "sustainable AI"
      ]
    }
  ]
}