executable-code

The paper proposes credit-addressable reasoning with executable code traces and localized reinforcement learning to enhance multimodal geometry reasoning, achieving significant accuracy improvements over baseline models like Qwen3-VL-8B.

The paper introduces Code World Model, a framework that combines language models for reasoning and coding with video models for visual rendering to simulate persistent world evolution in an open-ended manner.

RepoExec is a new benchmark for evaluating repository-level code generation focused on executability, functional correctness via automated test generation, and cross-file dependency integration. Experiments reveal that pretrained LLMs beat instruction-tuned ones on correctness, while fine-tuned models better leverage provided code dependencies.