VETR: Execution-Grounded Self-Improving Agent for Symbolic Graphics Reasoning
arXiv:2610.13098v1 Announce Type: new Abstract: Visual agents for symbolic graphics reasoning depend on executable world representations, yet continued improvement remains bottlenecked by ambiguous supervision from final renders. Recent agentic systems refine scenes through repeated interaction, but they lack mechanisms to recursively extract actionable evidence from execution traces, overlooking which proposed symbolic edits survived, disappeared, or interfered during program execution. To add…