Selection-Safe Conformal Early Exit: Distribution-Free Coverage under Adaptive Depth

Key points

• We construct one terminal path score from prefix maxima, calibrate a single split-conformal quantile, and return nested prefix sets that all contain the valid terminal set.1
• The guarantee is selection-safe without an alpha/L correction and holds simultaneously for every post-calibration label-free stopping policy that preserves terminal-set containment.1
• Across eight encoder task-backbone cells, pathwise selection attains macro coverage 0.901 versus 0.814 for naive layerwise selection, with average set size 1.41 and latency fraction 0.71.1
• In a joint exit-count/calibration-size sweep at n=100, pathwise set size moves only from 3.02 to 3.11 as exits increase from 4 to 24, while Bonferroni degenerates to the full label space for 12 or more exits.1

Sources (1)

• [1]Selection-Safe Conformal Early Exit: Distribution-Free Coverage under Adaptive Depth
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 06:22 PM

  Calibrating a conformal prediction set at each early exit does not preserve coverage once a data-dependent exit is selected.

  We construct one terminal path score from prefix maxima, calibrate a single split-conformal quantile, and return nested prefix sets that all contain the valid terminal set.

Extractive summary: sentences quoted from the sources.