Search papers, labs, and topics across Lattice.
This paper introduces a co-evolving Propose-then-Critic framework for GUI grounding, where a proposer generates pixel coordinates and a critic evaluates the proposals rendered on the screenshot. They use a maturity-aware adaptive co-evolutionary reinforcement learning paradigm to dynamically balance the training of the proposer and critic. Experiments across six benchmarks demonstrate significant improvements in grounding accuracy and critic reliability compared to static self-consistency strategies.
Learnable critics that evaluate the model's own GUI grounding proposals, rather than relying on static geometric heuristics, unlock substantial gains in accuracy.
Graphical User Interface (GUI) grounding requires mapping natural language instructions to precise pixel coordinates. However, due to visually homogeneous elements and dense layouts, models typically grasp semantic intent yet struggle with achieving precise localization. While scaling sampling attempts (Pass@k) reveals potential gains, static self-consistency strategies derived from geometric clustering often yield limited improvements, as the model's predictions tend to be spatially dispersed. In this paper, we propose replacing static consistency strategies with a learnable selection mechanism that selects the optimal target by critiquing its own proposals rendered on the screenshot. Given the significant disparity between the model's grounding and critiquing capabilities, we propose a co-evolving Propose-then-Critic framework. To jointly optimize these, we introduce a maturity-aware adaptive co-evolutionary reinforcement learning paradigm. This approach dynamically balances the training objectives of proposer and critic, where the diversity of the proposer's outputs enhances critic robustness, while the critic's maturing discrimination capability conversely unlocks the proposer's potential for extensive spatial exploration, fostering the mutual reinforcement and co-evolution of both capabilities, thereby ensuring generalizability to adapt to diverse and complex interface layouts. Extensive experiments over 6 benchmarks show that our method significantly enhances both grounding accuracy and critic reliability.