# How Reliable Are Automatic Evaluation Methods for Instruction-Tuned LLMs?

arXiv: https://arxiv.org/abs/2402.10770 · HTML: https://mlsanity.com/paper/2402.10770/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 49 references that are in the map.
- First posted on arXiv 16 Feb 2024.

## Builds on (most cited first, 25 of 25)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Finetuned Language Models Are Zero-Shot Learners](https://mlsanity.com/paper/2109.01652/) (2021)
- [Self-Instruct: Aligning Language Models with Self-Generated Instructions](https://mlsanity.com/paper/2212.10560/) (2022)
- [LIMA: Less Is More for Alignment](https://mlsanity.com/paper/2305.11206/) (2023)
- [ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks](https://mlsanity.com/paper/2303.15056/) (2023)
- [Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks](https://mlsanity.com/paper/2204.07705/) (2022)
- [ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate](https://mlsanity.com/paper/2308.07201/) (2023)
- [Cross-Task Generalization via Natural Language Crowdsourcing Instructions](https://mlsanity.com/paper/2104.08773/) (2021)
- [Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor](https://mlsanity.com/paper/2212.09689/) (2022)
- [ChatGPT Beyond English: Towards a Comprehensive Evaluation of Large Language Models in Multilingual Learning](https://mlsanity.com/paper/2304.05613/) (2023)
- [BLOOM+1: Adding Language Support to BLOOM for Zero-Shot Prompting](https://mlsanity.com/paper/2212.09535/) (2022)
- [Are Large Language Model-based Evaluators the Solution to Scaling Up Multilingual Evaluation?](https://mlsanity.com/paper/2309.07462/) (2023)
- [The Turking Test: Can Language Models Understand Instructions?](https://mlsanity.com/paper/2010.11982/) (2020)
- [Bactrian-X: Multilingual Replicable Instruction-Following Models with Low-Rank Adaptation](https://mlsanity.com/paper/2305.15011/) (2023)
- [Learning from Task Descriptions](https://mlsanity.com/paper/2011.08115/) (2020)
- [Ties Matter: Meta-Evaluating Modern Metrics with Pairwise Accuracy and Tie Calibration](https://mlsanity.com/paper/2305.14324/) (2023)
- [Help me write a poem: Instruction Tuning as a Vehicle for Collaborative Poetry Writing](https://mlsanity.com/paper/2210.13669/) (2022)
- [Monolingual or Multilingual Instruction Tuning: Which Makes a Better Alpaca](https://mlsanity.com/paper/2309.08958/) (2023)
- [Did You Read the Instructions? Rethinking the Effectiveness of Task Definitions in Instruction Learning](https://mlsanity.com/paper/2306.01150/) (2023)
- [Turning English-centric LLMs Into Polyglots: How Much Multilinguality Is Needed?](https://mlsanity.com/paper/2312.12683/) (2023)
- [Empowering Cross-lingual Abilities of Instruction-tuned Large Language Models by Translation-following demonstrations](https://mlsanity.com/paper/2308.14186/) (2023)
- [GPT-SW3: An Autoregressive Language Model for the Nordic Languages](https://mlsanity.com/paper/2305.12987/) (2023)

## Built on by (most cited first, 13 of 13)

- [Grade Like a Human: Rethinking Automated Assessment with Large Language Models](https://mlsanity.com/paper/2405.19694/) (2024)
- [Beyond the Surface: Enhancing LLM-as-a-Judge Alignment with Human via Internal Representations](https://mlsanity.com/paper/2508.03550/) (2025)
- [What Makes a Good Natural Language Prompt?](https://mlsanity.com/paper/2506.06950/) (2025)
- [CLEV: LLM-Based Evaluation Through Lightweight Efficient Voting for Free-Form Question-Answering](https://mlsanity.com/paper/2503.08542/) (2025)
- [Identifying Reliable Evaluation Metrics for Scientific Text Revision](https://mlsanity.com/paper/2506.04772/) (2025)
- [Characterizing Multimodal Long-form Summarization: A Case Study on Financial Reports](https://mlsanity.com/paper/2404.06162/) (2024)
- [Graders should cheat: privileged information enables expert-level automated evaluations](https://mlsanity.com/paper/2502.10961/) (2025)
- [Rapid Word Learning Through Meta In-Context Learning](https://mlsanity.com/paper/2502.14791/) (2025)
- [HalluScore: Large Language Model Hallucination Question Answering Benchmark](https://mlsanity.com/paper/2605.17007/) (2026)
- [CoKe: Customizable Fine-Grained Story Evaluation via Chain-of-Keyword Rationalization](https://mlsanity.com/paper/2503.17136/) (2025)
- [*-PLUIE: Personalisable metric with Llm Used for Improved Evaluation](https://mlsanity.com/paper/2602.15778/) (2026)
- [Leveraging Hierarchical Organization for Medical Multi-document Summarization](https://mlsanity.com/paper/2510.23104/) (2025)
- [MTQ-Eval: Multilingual Text Quality Evaluation for Language Models](https://mlsanity.com/paper/2511.09374/) (2025)

## Abstract (arXiv)

Work on instruction-tuned Large Language Models (LLMs) has used automatic methods based on text overlap and LLM judgments as cost-effective alternatives to human evaluation. In this paper, we perform a meta-evaluation of such methods and assess their reliability across a broad range of tasks. In evaluating how well automatic methods align with human evaluations, correlation metrics are the most commonly employed method despite their inherent limitations when dealing with ties and different scales. To address these shortcomings, we use Pairwise Accuracy as an alternative to standard correlation measures. We observe that while automatic evaluation methods can approximate human ratings under specific conditions, their validity is highly context-dependent. Specifically, the simple ROUGE-L metric correlates very well with human ratings for short-answer English tasks but is unreliable in free-form generation tasks and cross-lingual scenarios. The effectiveness of the more advanced method of using GPT-4 as a judge diminishes significantly if reference answers are not included in the prompt, which is the scenario where this method has the potential to provide the most value compared to other metrics. Our findings enhance the understanding of how automatic methods should be applied and interpreted when developing and evaluating instruction-tuned LLMs.