FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ye, Seonghyeon, Kim, Doyoung, Kim, Sungdong, Hwang, Hyeonbin, Kim, Seungone, Jo, Yongrae, Thorne, James, Kim, Juho, Seo, Minjoon
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910408600715264
author Ye, Seonghyeon
Kim, Doyoung
Kim, Sungdong
Hwang, Hyeonbin
Kim, Seungone
Jo, Yongrae
Thorne, James
Kim, Juho
Seo, Minjoon
author_facet Ye, Seonghyeon
Kim, Doyoung
Kim, Sungdong
Hwang, Hyeonbin
Kim, Seungone
Jo, Yongrae
Thorne, James
Kim, Juho
Seo, Minjoon
contents Evaluation of Large Language Models (LLMs) is challenging because instruction-following necessitates alignment with human values and the required set of skills varies depending on the instruction. However, previous studies have mainly focused on coarse-grained evaluation (i.e. overall preference-based evaluation), which limits interpretability since it does not consider the nature of user instructions that require instance-wise skill composition. In this paper, we introduce FLASK (Fine-grained Language Model Evaluation based on Alignment Skill Sets), a fine-grained evaluation protocol for both human-based and model-based evaluation which decomposes coarse-level scoring to a skill set-level scoring for each instruction. We experimentally observe that the fine-graininess of evaluation is crucial for attaining a holistic view of model performance and increasing the reliability of the evaluation. Using FLASK, we compare multiple open-source and proprietary LLMs and observe a high correlation between model-based and human-based evaluations. We publicly release the evaluation data and code implementation at https://github.com/kaistAI/FLASK.
format Preprint
id arxiv_https___arxiv_org_abs_2307_10928
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets
Ye, Seonghyeon
Kim, Doyoung
Kim, Sungdong
Hwang, Hyeonbin
Kim, Seungone
Jo, Yongrae
Thorne, James
Kim, Juho
Seo, Minjoon
Computation and Language
Artificial Intelligence
Evaluation of Large Language Models (LLMs) is challenging because instruction-following necessitates alignment with human values and the required set of skills varies depending on the instruction. However, previous studies have mainly focused on coarse-grained evaluation (i.e. overall preference-based evaluation), which limits interpretability since it does not consider the nature of user instructions that require instance-wise skill composition. In this paper, we introduce FLASK (Fine-grained Language Model Evaluation based on Alignment Skill Sets), a fine-grained evaluation protocol for both human-based and model-based evaluation which decomposes coarse-level scoring to a skill set-level scoring for each instruction. We experimentally observe that the fine-graininess of evaluation is crucial for attaining a holistic view of model performance and increasing the reliability of the evaluation. Using FLASK, we compare multiple open-source and proprietary LLMs and observe a high correlation between model-based and human-based evaluations. We publicly release the evaluation data and code implementation at https://github.com/kaistAI/FLASK.
title FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2307.10928