Evaluating Large Language Models on Controlled Generation Tasks
This addresses the problem of controllability in language generation for AI researchers, but it is incremental as it builds on existing benchmark analyses.
The study evaluated large language models on controlled generation tasks, finding that they struggle to meet fine-grained hard constraints compared to state-of-the-art fine-tuned smaller models, with results showing a spectrum from falling behind to exceeding in some cases.
While recent studies have looked into the abilities of large language models in various benchmark tasks, including question generation, reading comprehension, multilingual and etc, there have been few studies looking into the controllability of large language models on generation tasks. We present an extensive analysis of various benchmarks including a sentence planning benchmark with different granularities. After comparing large language models against state-of-the-start finetuned smaller models, we present a spectrum showing large language models falling behind, are comparable, or exceed the ability of smaller models. We conclude that **large language models struggle at meeting fine-grained hard constraints**.