Value-based Search in Execution Space for Mapping Instructions to Programs
This work addresses a bottleneck in program synthesis from instructions, offering a domain-specific improvement for tasks like the SCONE dataset.
The paper tackles the challenge of mapping natural language instructions to programs when instructions grow long, by proposing a search algorithm that uses target world states to train a critic network for scoring search states. The result shows dramatic performance improvements on the SCONE dataset across all three domains compared to standard beam search and other baselines.
Training models to map natural language instructions to programs given target world supervision only requires searching for good programs at training time. Search is commonly done using beam search in the space of partial programs or program trees, but as the length of the instructions grows finding a good program becomes difficult. In this work, we propose a search algorithm that uses the target world state, known at training time, to train a critic network that predicts the expected reward of every search state. We then score search states on the beam by interpolating their expected reward with the likelihood of programs represented by the search state. Moreover, we search not in the space of programs but in a more compressed state of program executions, augmented with recent entities and actions. On the SCONE dataset, we show that our algorithm dramatically improves performance on all three domains compared to standard beam search and other baselines.