Prompt-efficient proposal generation

Develop prompt-efficient proposal generation for the Cue-Calibrated Linguistic Object Onboarding front-end to reduce the computational cost of processing multiple linguistic prompts with SAM 3 while preserving target-object proposal quality.

Background

Cue-Calibrated Linguistic Object Onboarding uses multiple linguistic memories as prompts for SAM 3. This design improves proposal quality but can incur substantial detection latency when many objects are onboarded, because SAM 3 must process multiple target-prompt pairs. The paper therefore identifies prompt-efficient proposal generation as a concrete direction for future work aimed at alleviating this runtime bottleneck.

References

Finally, CLON is not optimized for worst-case runtime: SAM 3 must process multiple prompts, and proposal-object scoring scales with the number of onboarded objects. Future work will explore prompt-efficient proposal generation and stronger object-set reliability estimation.

— CLON: Cue-Calibrated Linguistic Object Onboarding for Zero-Shot 6D Pose Front-Ends  (2609.04784 - Ji et al., 4 Sep 2026) in Section 5, Limitations