A Systematic Study and Comprehensive Evaluation of ChatGPT on Benchmark Datasets

Sep 17, 2026

Uploaded by Dipper

Large language models like ChatGPT have attracted broad interest, yet their evaluation on benchmark academic datasets remains limited because generative outputs must be compared against ground truth. This paper presents a thorough assessment of ChatGPT across diverse academic tasks, including question answering, summarization, code generation, reasoning, math, translation, bias detection, and ethical considerations. The study evaluates 140 tasks and analyzes 255K generated responses, identifying strengths, weaknesses, and emergent multi-query instruction following capabilities, while motivating targeted real-world deployment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages37

Reading Time1 h 33 min

AccessFree to download

Abstract

1 Introduction

A Systematic Study and Comprehensive Evaluation of ChatGPT on Benchmark Datasets Md Tahmid Rahman Laskar∗† §, M Saiful Bari∗‡, Mizanur Rahman∗† ¶ Md Amran Hossen Bhuiyan†, Shafiq Joty‡$ , Jimmy Xiangji Huang††York University,‡Nanyang Technological University, §Dialpad Canada Inc., ¶Royal Bank of Canada, $ Salesforce Research {tahmid20,mizanurr,amran,[jhuang}@yorku.ca](jhuang}@yorku.ca)[ ](jhuang}@yorku.ca){bari0001,[srjoty}@ntu.edu.sg](srjoty}@ntu.edu.sg) arXiv :2305 . 18486v4 [ cs .CL] 5 Jul 2023 Abstract The development of large language models (LLMs) such as ChatGPT 1 has brought a lot of attention recently. However, their evaluation in the benchmark academic datasets remains under-explored due to the difficulty of evaluating the generative outputs produced by this model against the ground truth. In this paper, we aim to present a thorough evaluation of ChatGPT’s performance on diverse academic datasets, covering tasks like questionanswering, text summarization, code generation, commonsense reasoning, mathematical problem-solving, machine translation, bias detection, and ethical considerations. Specifically, we evaluate ChatGPT across 140 tasks and analyze 255K responses it generates in these datasets. This makes our work the largest evaluation of ChatGPT in NLP benchmarks. In short, our study aims to validate the strengthsand weaknesses of ChatGPT in various tasks and provide insights for future research using LLMs. We also report a new emergent ability to follow multi-query instructions that we mostly found in ChatGPT and other instruction-tuned models. Our extensive evaluation shows that even though ChatGPT is capable of performing a wide variety of tasks, and may obtain impressive performance in several benchmark datasets, it is still far from achieving the ability to reliably solve many challenging tasks. By providing a thorough assessment of ChatGPT’s performance across diverse NLP tasks, this paper sets the stage for a targeted deployment of ChatGPT-like LLMs in real-world applications. 1 Introduction In recent years, the introduction of transformerbased (Vaswani et al., 2017) language models (LMs), such as BERT (Devlin et al., 2018), GPT (Radford et al., 2019), T5 (Raffel et al., 2020), etc. have led to significant advancements in NLP (Liu ∗ First three authors contributed equally. 1 [https://chat.openai.com/](https://chat.openai.com/) et al., 2019 ; Sanh et al., 2019 ; Lan et al., 2019 ; Lewis et al., 2020 ; Clark et al., 2020) . The effectiveness of these models was evaluated by finetuning them on benchmark datasets (Wang et al., 2018, 2019), achieving state-of-the-art (SOTA) performance across various tasks. Recently, large language models (LLMs) such as GPT-3 (Brown et al., 2020) have demonstrated in-context-learning capability without requiring any fine-tuning on taskspecific data. The impressive performance of GPT- 3 and other LLMs (Scao et al., 2022 ; Tay et al., 2022 ; Thoppilan et al., 2022 ; Fedus et al., 2021 ; Hoffmann et al., 2022 ; Zeng et al., 2022) in fewshot learning scenarios is a major finding as this helps LLMs to be more efficient, making it possible to use LM-as-a-service (Sun et al., 2022) to empower a set of new real-world applications. Intuitively, in-context learning works by learning through analogies drawn from the given demonstration examples (Dong et al., 2023) . After a largescale pre-training with a self-supervision objective, LLMs can identify task-level prior patterns from the given prompt and generate a relevant continuation. Large-scale pretraining also helps them to acquire emergent capabilities like Chain of Thought (Wei et al., 2022a) . However, training only with self-supervision lacks grounding to real-world concepts and may not align well with its inference-time use cases resulting in unhelpful, hallucinated and sometimes toxic output (Ouyang et al., 2022) . Thus, instead of learning meta-tasks in an implicit way from raw texts, recent approache

FAQs

Why is evaluating ChatGPT on benchmark datasets difficult?

Because generative outputs must be evaluated against ground truth, which is non-trivial for natural language generation tasks.

What kinds of tasks does the paper evaluate for ChatGPT?

It covers question answering, text summarization, code generation, commonsense reasoning, mathematical problem-solving, machine translation, bias detection, and ethical considerations.

How extensive is the evaluation conducted in the paper?

The paper evaluates ChatGPT across 140 tasks and analyzes 255K responses generated for these benchmark datasets.

Sep 17, 2026

Uploaded by Dipper