# EVA2.0: Investigating Open-Domain Chinese Dialogue Systems with Large-Scale Pre-Training

Source: [EVA2.0: Investigating Open-Domain Chinese Dialogue Systems with Large-Scale Pre-Training](<https://www.sota2.com/research/paper/68fa70f41e209131cea4e5ca>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-27T06:27:18.233Z

Authors: Yuxian Gu, Jiaxin Wen, Hao Sun, Yi Song, Pei Ke, Chujie Zheng, Zheng Zhang, Jianzhu Yao, Lei Liu, Xiaoyan Zhu, Minlie Huang

Published: 2022-03-17T13:33:17.000Z

## Abstract

Large-scale pre-training has shown remarkable performance in building open-domain dialogue systems. However, previous works mainly focus on showing and evaluating the conversational performance of the released dialogue model, ignoring the discussion of some key factors towards a powerful human-like chatbot, especially in Chinese scenarios. In this paper, we conduct extensive experiments to investigate these under-explored factors, including data quality control, model architecture designs, training approaches, and decoding strategies. We propose EVA2.0, a large-scale pre-trained open-domain Chinese dialogue model with 2.8 billion parameters, and will make our models and codes publicly available. Automatic and human evaluations show that EVA2.0 significantly outperforms other open-source counterparts. We also discuss the limitations of this work by presenting some failure cases and pose some future research directions on large-scale Chinese open-domain dialogue systems.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Short-text Generation on Weibo (test)](<https://www.sota2.com/research/sota/short-text-generation-on-weibo-test.md>): Short-text generation; Weibo (test); F1 Score: 12.94.

- [Short-text Generation on LCCC (test)](<https://www.sota2.com/research/sota/short-text-generation-on-lccc-test.md>): Short-text generation; LCCC (test); F1 Score: 11.75.

- [Open-domain Conversation on Chinese open-domain conversation Self-chat (test)](<https://www.sota2.com/research/sota/open-domain-conversation-on-chinese-open-domain-conversation-self-chat-test.md>): Open-domain Conversation; Chinese open-domain conversation Self-chat (test); Coherence: 150.8.

- [Short-text Generation on Douban (test)](<https://www.sota2.com/research/sota/short-text-generation-on-douban-test.md>): Short-text generation; Douban (test); F1 Score: 9.59.

- [Short-text generation on Douban](<https://www.sota2.com/research/sota/short-text-generation-on-douban.md>): Short-text generation; Douban; Informativeness: 2.5.

- [Short-text Generation on Weibo](<https://www.sota2.com/research/sota/short-text-generation-on-weibo.md>): Short-text generation; Weibo; Informativeness: 2.75.

- [Short-text generation on LCCC](<https://www.sota2.com/research/sota/short-text-generation-on-lccc.md>): Short-text generation; LCCC; Informativeness: 2.83.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2203.09313>)

[PDF](<https://arxiv.org/pdf/2203.09313.pdf>)