Submit Paper

FMAI welcomes submissions that make agent failures reproducible, measurable, and fixable. We are especially interested in work that turns messy agent behavior into concrete research assets: precise failure definitions, minimal triggers, actionable diagnostics, and verified interventions.

• Operational definitions with clear failure boundaries.
• Reproducible triggers or minimal reproductions for the failure.
• Comparable evaluation protocols or trace-level diagnostics.
• Mitigation, repair, or intervention strategies with explicit evidence about what improves and what does not.

We also strongly encourage well-documented negative results. Failed mitigations, brittle training tricks, and interventions that only work under narrow conditions are all in scope if the analysis is careful and the lesson is transferable.