Controllable Text Generation in the Instruction-Tuning Era
This work addresses the need for research on controllable text generation specifically for instruction-tuned models, which is incremental as it adapts existing paradigms to a new context.
The paper tackles the problem of controllable text generation with instruction-tuned language models by benchmarking methods on a new testbed, finding that prompting-based approaches outperform specialized methods on most tasks, matching human performance on stylistic tasks but lagging on structural ones.
While most research on controllable text generation has focused on steering base Language Models, the emerging instruction-tuning and prompting paradigm offers an alternate approach to controllability. We compile and release ConGenBench, a testbed of 17 different controllable generation tasks, using a subset of it to benchmark the performance of 9 different baselines and methods on Instruction-tuned Language Models. To our surprise, we find that prompting-based approaches outperform controllable text generation methods on most datasets and tasks, highlighting a need for research on controllable text generation with Instruction-tuned Language Models in specific. Prompt-based approaches match human performance on most stylistic tasks while lagging on structural tasks, foregrounding a need to study more varied constraints and more challenging stylistic tasks. To facilitate such research, we provide an algorithm that uses only a task dataset and a Large Language Model with in-context capabilities to automatically generate a constraint dataset. This method eliminates the fields dependence on pre-curated constraint datasets, hence vastly expanding the range of constraints that can be studied in the future.