Revisiting Out-of-distribution Robustness in NLP: Benchmark, Analysis, and LLMs Evaluations
This work addresses the need for more rigorous evaluation of OOD robustness in NLP, which is crucial for developing reliable models in real-world applications, though it is incremental as it builds on existing benchmarks and methods.
This paper tackles the problem of evaluating out-of-distribution (OOD) robustness in NLP by proposing BOSS, a benchmark suite with challenging distribution shifts across 5 tasks and 20 datasets, and finds that fine-tuned small models outperform large language models (LLMs) on in-distribution data but LLMs with in-context learning are better for OOD instances, with no significant improvement from classic OOD methods over vanilla fine-tuning.
This paper reexamines the research on out-of-distribution (OOD) robustness in the field of NLP. We find that the distribution shift settings in previous studies commonly lack adequate challenges, hindering the accurate evaluation of OOD robustness. To address these issues, we propose a benchmark construction protocol that ensures clear differentiation and challenging distribution shifts. Then we introduce BOSS, a Benchmark suite for Out-of-distribution robustneSS evaluation covering 5 tasks and 20 datasets. Based on BOSS, we conduct a series of experiments on pre-trained language models for analysis and evaluation of OOD robustness. First, for vanilla fine-tuning, we examine the relationship between in-distribution (ID) and OOD performance. We identify three typical types that unveil the inner learning mechanism, which could potentially facilitate the forecasting of OOD robustness, correlating with the advancements on ID datasets. Then, we evaluate 5 classic methods on BOSS and find that, despite exhibiting some effectiveness in specific cases, they do not offer significant improvement compared to vanilla fine-tuning. Further, we evaluate 5 LLMs with various adaptation paradigms and find that when sufficient ID data is available, fine-tuning domain-specific models outperform LLMs on ID examples significantly. However, in the case of OOD instances, prioritizing LLMs with in-context learning yields better results. We identify that both fine-tuned small models and LLMs face challenges in effectively addressing downstream tasks. The code is public at \url{https://github.com/lifan-yuan/OOD_NLP}.