InfiR : Crafting Effective Small Language Models and Multimodal Small Language Models in Reasoning

TL;DR

This paper presents InfiR, a training pipeline for small and multimodal small language models that achieve strong reasoning performance, are resource-efficient, and suitable for edge deployment, addressing high costs and privacy issues of large models.

Contribution

The paper introduces a novel training pipeline for small and multimodal language models that enhances reasoning and enables efficient deployment on edge devices.

Findings

Achieves state-of-the-art reasoning performance among small models

Reduces computational and privacy concerns compared to large models

Facilitates deployment on resource-constrained devices

Abstract

Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) have made significant advancements in reasoning capabilities. However, they still face challenges such as high computational demands and privacy concerns. This paper focuses on developing efficient Small Language Models (SLMs) and Multimodal Small Language Models (MSLMs) that retain competitive reasoning abilities. We introduce a novel training pipeline that enhances reasoning capabilities and facilitates deployment on edge devices, achieving state-of-the-art performance while minimizing development costs. \InfR~ aims to advance AI systems by improving reasoning, reducing adoption barriers, and addressing privacy concerns through smaller model sizes. Resources are available at https://github. com/Reallm-Labs/InfiR.