The Vulnerability of Language Model Benchmarks: Do They Accurately Reflect True LLM Performance?
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_ | 1866912144888430592 |
|---|---|
| author | Banerjee, Sourav Agarwal, Ayushi Singh, Eishkaran |
| author_facet | Banerjee, Sourav Agarwal, Ayushi Singh, Eishkaran |
| contents | The pursuit of leaderboard rankings in Large Language Models (LLMs) has created a fundamental paradox: models excel at standardized tests while failing to demonstrate genuine language understanding and adaptability. Our systematic analysis of NLP evaluation frameworks reveals pervasive vulnerabilities across the evaluation spectrum, from basic metrics to complex benchmarks like GLUE and MMLU. These vulnerabilities manifest through benchmark exploitation, dataset contamination, and evaluation bias, creating a false perception of progress in language understanding capabilities. Through extensive review of contemporary evaluation approaches, we identify significant limitations in static benchmark designs, human evaluation protocols, and LLM-as-judge frameworks, all of which compromise the reliability of current performance assessments. As LLM capabilities evolve and existing benchmarks become redundant, we lay the groundwork for new evaluation methods that resist manipulation, minimize data contamination, and assess domain-specific tasks. This requires frameworks that are adapted dynamically, addressing current limitations and providing a more accurate reflection of LLM performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_03597 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | The Vulnerability of Language Model Benchmarks: Do They Accurately Reflect True LLM Performance? Banerjee, Sourav Agarwal, Ayushi Singh, Eishkaran Computation and Language Machine Learning The pursuit of leaderboard rankings in Large Language Models (LLMs) has created a fundamental paradox: models excel at standardized tests while failing to demonstrate genuine language understanding and adaptability. Our systematic analysis of NLP evaluation frameworks reveals pervasive vulnerabilities across the evaluation spectrum, from basic metrics to complex benchmarks like GLUE and MMLU. These vulnerabilities manifest through benchmark exploitation, dataset contamination, and evaluation bias, creating a false perception of progress in language understanding capabilities. Through extensive review of contemporary evaluation approaches, we identify significant limitations in static benchmark designs, human evaluation protocols, and LLM-as-judge frameworks, all of which compromise the reliability of current performance assessments. As LLM capabilities evolve and existing benchmarks become redundant, we lay the groundwork for new evaluation methods that resist manipulation, minimize data contamination, and assess domain-specific tasks. This requires frameworks that are adapted dynamically, addressing current limitations and providing a more accurate reflection of LLM performance. |
| title | The Vulnerability of Language Model Benchmarks: Do They Accurately Reflect True LLM Performance? |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2412.03597 |