TeleChat Technical Report

TL;DR

TeleChat introduces a series of large language models with 3B, 7B, and 12B parameters, trained on diverse multilingual data, fine-tuned for human preferences, and evaluated across multiple tasks, with models and resources released publicly.

Contribution

The paper presents a new set of multilingual LLMs with detailed training and fine-tuning methodology, and releases models and data for community use.

Findings

TeleChat models perform comparably to similar-sized open-source models.

Models are effective across language understanding, reasoning, and code generation tasks.

Public release of models and resources supports further research.

Abstract

In this technical report, we present TeleChat, a collection of large language models (LLMs) with parameters of 3 billion, 7 billion and 12 billion. It includes pretrained language models as well as fine-tuned chat models that is aligned with human preferences. TeleChat is initially pretrained on an extensive corpus containing a diverse collection of texts from both English and Chinese languages, including trillions of tokens. Subsequently, the model undergoes fine-tuning to align with human preferences, following a detailed methodology that we describe. We evaluate the performance of TeleChat on various tasks, including language understanding, mathematics, reasoning, code generation, and knowledge-based question answering. Our findings indicate that TeleChat achieves comparable performance to other open-source models of similar size across a wide range of public benchmarks. To support future research and applications utilizing LLMs, we release the fine-tuned model checkpoints of TeleChat's 7B and 12B variant, along with code and a portion of our pretraining data, to the public community.