Evaluating LLMs' Mathematical and Coding Competency through Ontology-guided Interventions
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866929572671389696 |
|---|---|
| author | Hong, Pengfei Majumder, Navonil Ghosal, Deepanway Aditya, Somak Mihalcea, Rada Poria, Soujanya |
| author_facet | Hong, Pengfei Majumder, Navonil Ghosal, Deepanway Aditya, Somak Mihalcea, Rada Poria, Soujanya |
| contents | Recent advancements in Large Language Models (LLMs) have showcased striking results on existing logical reasoning benchmarks, with some models even surpassing human performance. However, the true depth of their competencies and robustness in reasoning tasks remains an open question. To this end, in this paper, we focus on two popular reasoning tasks: arithmetic reasoning and code generation. Particularly, we introduce (i) a general ontology of perturbations for math and coding questions, (ii) a semi-automatic method to apply these perturbations, and (iii) two datasets, GSMORE and HUMANEVAL-CORE, respectively, of perturbed math and coding problems to probe LLM capabilities in numeric reasoning and coding tasks. Through comprehensive evaluations of both closed-source and open-source LLMs, we show a significant performance drop across all the models against the perturbed questions, suggesting that the current LLMs lack robust problem solving skills and structured reasoning abilities in many areas, as defined by our ontology. We open-source the datasets and source codes at: https://github.com/declare-lab/LLM-ReasoningTest. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_09395 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Evaluating LLMs' Mathematical and Coding Competency through Ontology-guided Interventions Hong, Pengfei Majumder, Navonil Ghosal, Deepanway Aditya, Somak Mihalcea, Rada Poria, Soujanya Computation and Language Recent advancements in Large Language Models (LLMs) have showcased striking results on existing logical reasoning benchmarks, with some models even surpassing human performance. However, the true depth of their competencies and robustness in reasoning tasks remains an open question. To this end, in this paper, we focus on two popular reasoning tasks: arithmetic reasoning and code generation. Particularly, we introduce (i) a general ontology of perturbations for math and coding questions, (ii) a semi-automatic method to apply these perturbations, and (iii) two datasets, GSMORE and HUMANEVAL-CORE, respectively, of perturbed math and coding problems to probe LLM capabilities in numeric reasoning and coding tasks. Through comprehensive evaluations of both closed-source and open-source LLMs, we show a significant performance drop across all the models against the perturbed questions, suggesting that the current LLMs lack robust problem solving skills and structured reasoning abilities in many areas, as defined by our ontology. We open-source the datasets and source codes at: https://github.com/declare-lab/LLM-ReasoningTest. |
| title | Evaluating LLMs' Mathematical and Coding Competency through Ontology-guided Interventions |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2401.09395 |