mlsanityData to 7 Oct 2026

Task-Specific Efficiency Analysis: When Small Language Models Outperform Large Language Models

arXiv 2603.21389 ↗First posted 22 Mar 2026cs.CL cs.LGcited by 15

Not on the map. Why?

Abstract (arXiv)

Large Language Models achieve remarkable performance but incur substantial computational costs unsuitable for resource-constrained deployments. This paper presents the first comprehensive task-specific efficiency analysis comparing 16 language models across five diverse NLP tasks. We introduce the Performance-Efficiency Ratio (PER), a novel metric integrating accuracy, throughput, memory, and latency through geometric mean normalization. Our systematic evaluation reveals that small models (0.5--3B parameters) achieve superior PER scores across all given tasks. These findings establish quantitative foundations for deploying small models in production environments prioritizing inference efficiency over marginal accuracy gains.

Key facts. Cited by 15 papers in the map (data through 7 Oct 2026). Builds on 41 of its 65 references that are in the map. First posted on arXiv 22 Mar 2026.

Builds on

41 of 65 references.

References · 41

  1. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  2. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
  3. 2019HellaSwag: Can a Machine Really Finish Your Sentence?cited by 4,018
  4. 2018Know What You Don't Know: Unanswerable Questions for SQuADcited by 1,538
  5. 2024MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategiescited by 694
  6. 2020It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learnerscited by 668
  7. 2024A Comprehensive Survey of Small Language Models in the Era of Large Language Models: Techniques, Enhancements, Applications, Collaboration with LLMs, and Trustworthinesscited by 125
  8. 2025OstQuant: Refining Large Language Model Quantization with Orthogonal and Scaling Transformations for Better Distribution Fittingcited by 94
  9. 2025VL-JEPA: Joint Embedding Predictive Architecture for Vision-languagecited by 61
  10. 2025FineCIR: Explicit Parsing of Fine-Grained Modification Semantics for Composed Image Retrievalcited by 41
  11. 2025VisRL: Intention-Driven Visual Perception via Reinforced Reasoningcited by 39
  12. 2025MQuant: Unleashing the Inference Potential of Multimodal Large Language Models via Full Static Quantizationcited by 35
29 more
  1. 2022BALF: Simple and Efficient Blur Aware Local Feature Detectorcited by 34
  2. 2024Human Motion Instruction Tuningcited by 29
  3. 2025What Is The Best 3D Scene Representation for Robotics? From Geometric to Foundation Modelscited by 29
  4. 2025SIFThinker: Spatially-Aware Image Focus for Visual Reasoningcited by 25
  5. 2024SGLP: A Similarity Guided Fast Layer Partition Pruning for Compressing Large Deep Modelscited by 24
  6. 2023SyreaNet: A Physically Guided Underwater Image Enhancement Framework Integrating Synthetic and Real Imagescited by 23
  7. 2025Does Tone Change the Answer? Evaluating Prompt Politeness Effects on Modern LLMs: GPT, Gemini, and LLaMAcited by 23
  8. 2025GaussianDWM: 3D Gaussian Driving World Model for Unified Scene Understanding and Multi-Modal Generationcited by 23
  9. 2023FedLPA: One-shot Federated Learning with Layer-Wise Posterior Aggregationcited by 22
  10. 2024Mitigating the Risk of Health Inequity Exacerbated by Large Language Modelscited by 21
  11. 2025GSQ-Tuning: Group-Shared Exponents Integer in Fully Quantized Training for LLMs On-Device Fine-tuningcited by 21
  12. 2024SILMM: Self-Improving Large Multimodal Models for Compositional Text-to-Image Generationcited by 15
  13. 2024Learn from Global Correlations: Enhancing Evolutionary Algorithm via Spectral GNNcited by 15
  14. 2025ChatMotion: A Multimodal Multi-Agent for Human Motion Analysiscited by 13
  15. 2026DynaWeb: Model-Based Reinforcement Learning of Web Agentscited by 12
  16. 2025CausalSR: Structural Causal Model-Driven Super-Resolution with Counterfactual Inferencecited by 11
  17. 2024Are Small Language Models Ready to Compete with Large Language Models for Practical Applications?cited by 9
  18. 2024Contextual Bandits for Unbounded Context Distributionscited by 8
  19. 2025Assessing Small Language Models for Code Generation: An Empirical Study with Benchmarkscited by 8
  20. 2024Efficient Partitioning Vision Transformer on Edge Devices for Distributed Inferencecited by 6
  21. 2025Subject Information Extraction for Novelty Detection with Domain Shiftscited by 6
  22. 2025Reference-free Adversarial Sex Obfuscation in Speechcited by 6
  23. 2025Hallucination reduction with CASAL: Contrastive Activation Steering For Amortized Learningcited by 6
  24. 2025A Survey of Link Prediction in N-ary Knowledge Graphscited by 4
  25. 2025Urban-MAS: Human-Centered Urban Prediction with LLM-Based Multi-Agent Systemcited by 4
  26. 2026Adaptive traffic signal control optimization using a novel road partition and multi-channel state representation methodcited by 4
  27. 2025Differentiable NMS via Sinkhorn Matching for End-to-End Fabric Defect Detectioncited by 3
  28. 2023Few-shot Link Prediction on N-ary Factscited by 2
  29. 2026Fair-Gate: Fairness-Aware Interpretable Risk Gating for Sex-Fair Voice Biometricscited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 15.

15 papers.

Sort:

By area:

  1. 2026When Safe Skills Collide: Measuring Compositional Risk in Agent Skill Ecosystemscited by 14
  2. 2026Sustainable Hybrid Document-Routed Retrieval for Financial RAG: Resolving the Robustness-Precision Trade-offcited by 13
  3. 2026Bridging Reasoning Trajectories in On-Policy Distillation via Near-Future Guidancecited by 12
  4. 2026Cornerstones or Stumbling Blocks? Deciphering the Rock Tokens in On-Policy Distillationcited by 11
  5. 2026The Efficiency Frontier: A Unified Framework for Cost-Performance Optimization in LLM Context Managementcited by 6
  6. 2026MUSE: Resolving Manifold Misalignment in Visual Tokenization via Topological Orthogonalitycited by 3
  7. 2026Interpretable vs Learned Encoders for High-Cardinality Fraud Detectioncited by 2
  8. 2026Parametric Memory Decoding for Zero-Shot Routing in LoRA-Based External Parametric Memorycited by 2
  9. 2026Playing DOOM with 1.3M Parameters: Specialized Small Models vs Large Language Models for Real-Time Game Controlcited by 1
  10. 2026SynGR: Unleashing the Potential of Cross-Modal Synergy for Generative Recommendationcited by 1
  11. 2026Agora: Toward Autonomous Bug Detection in Production-Level Consensus Protocols with LLM Agentscited by 1
  12. 2026Energy-Efficient On-Device RAG on a Mobile NPU: System Design and Benchmark on Snapdragon X Elitecited by 1
3 more
  1. 2026OCC-RAG: Optimal Cognitive Core for Faithful Question Answeringcited by 0
  2. 2026Treatment Response Optimized Clinical Decision Support AI System via Digital Twin Simulationcited by 0
  3. 2026MARS: Multi-Agent Re-ranking for Repeat-Order Food Delivery Recommendationcited by 0