PLUGH: A Benchmark for Spatial Understanding and Reasoning in Large Language Models
Fuente:
arXiv
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913462936928256 |
|---|---|
| author | Tikhonov, Alexey |
| author_facet | Tikhonov, Alexey |
| contents | We present PLUGH (https://www.urbandictionary.com/define.php?term=plugh), a modern benchmark that currently consists of 5 tasks, each with 125 input texts extracted from 48 different games and representing 61 different (non-isomorphic) spatial graphs to assess the abilities of Large Language Models (LLMs) for spatial understanding and reasoning. Our evaluation of API-based and open-sourced LLMs shows that while some commercial LLMs exhibit strong reasoning abilities, open-sourced competitors can demonstrate almost the same level of quality; however, all models still have significant room for improvement. We identify typical reasons for LLM failures and discuss possible ways to deal with them. Datasets and evaluation code are released (https://github.com/altsoph/PLUGH). |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_04648 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | PLUGH: A Benchmark for Spatial Understanding and Reasoning in Large Language Models Tikhonov, Alexey Computation and Language Artificial Intelligence Information Retrieval 68T50, 68T20 I.2.7; I.2.8; G.2.2 We present PLUGH (https://www.urbandictionary.com/define.php?term=plugh), a modern benchmark that currently consists of 5 tasks, each with 125 input texts extracted from 48 different games and representing 61 different (non-isomorphic) spatial graphs to assess the abilities of Large Language Models (LLMs) for spatial understanding and reasoning. Our evaluation of API-based and open-sourced LLMs shows that while some commercial LLMs exhibit strong reasoning abilities, open-sourced competitors can demonstrate almost the same level of quality; however, all models still have significant room for improvement. We identify typical reasons for LLM failures and discuss possible ways to deal with them. Datasets and evaluation code are released (https://github.com/altsoph/PLUGH). |
| title | PLUGH: A Benchmark for Spatial Understanding and Reasoning in Large Language Models |
| topic | Computation and Language Artificial Intelligence Information Retrieval 68T50, 68T20 I.2.7; I.2.8; G.2.2 |
| url | https://arxiv.org/abs/2408.04648 |