lilGym: Natural Language Visual Reasoning with Reinforcement Learning

TL;DR

lilGym is a new benchmark for language-conditioned reinforcement learning in visual environments, featuring complex natural language statements and a novel reward computation method, posing a challenging problem for current models.

Contribution

We introduce lilGym, a benchmark with annotated executable programs for exact reward computation, enabling diverse and challenging language-visual RL tasks.

Findings

Existing models achieve limited performance on lilGym.

The benchmark reveals gaps in current reinforcement learning methods.

lilGym provides a challenging testbed for future research.

Abstract

We present lilGym, a new benchmark for language-conditioned reinforcement learning in visual environments. lilGym is based on 2,661 highly-compositional human-written natural language statements grounded in an interactive visual environment. We introduce a new approach for exact reward computation in every possible world state by annotating all statements with executable Python programs. Each statement is paired with multiple start states and reward functions to form thousands of distinct Markov Decision Processes of varying difficulty. We experiment with lilGym with different models and learning regimes. Our results and analysis show that while existing methods are able to achieve non-trivial performance, lilGym forms a challenging open problem. lilGym is available at https://lil.nlp.cornell.edu/lilgym/.