Across 0.8B-12B models plus a frontier arm, apparent self-correction effects are dominated by format-recovery and format-loss artifacts, with near-zero content-level change at capable scale.
Reflexion: Language agents with verbal reinforcement learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
The Calibration Floor: Format Repair Can Masquerade as Self-Correction at Small-to-Mid Scale
Across 0.8B-12B models plus a frontier arm, apparent self-correction effects are dominated by format-recovery and format-loss artifacts, with near-zero content-level change at capable scale.