Skip to main content
Search

# Cancellation-Aware Response Masking

1 related articles

Subscribe to updates
  1. arXiv

    CARM replaces geometric-mean masking with absolute-value averaging, raising mean@16 by up to 3.13 points on AIME 2024/2025/2026 and BeyondAIME

    The work proposes Cancellation-Aware Response Masking (CARM), a sequence-level mask that takes the absolute value of each token log-ratio before averaging so that opposing probability changes cannot cancel, and proves that accepted responses satisfy a joint bound on the fraction of sampled-token ratios outside a prescribed band and their mean log-distance beyond its boundaries; in experiments on mathematical reasoning and code generation, CARM improves mean@16 averaged over AIME 2024/2025/2026 and BeyondAIME by up to 3.13 percentage points over geometric-mean masking and increases average pass@1 across four code benchmarks by 2.88 points over the strongest evaluated baseline.

You’re all caught up