Md-Hakim/

Termination_Calibration_R3

This repository contains the full rollout data behind "Termination Calibration Recovers Answers That Reasoning Models Compute but Never Express." It captures model generations at three points in the causal pipeline — before training, after correctness-only RL (R1), and after termination-calibrated RL (R3) — across three DeepSeek-R1-distilled backbones, plus one additional composition condition (R3 applied on top of a DECS… See the full description on the dataset page: https://huggingface.co/datasets/Md-Hakim/Termination_Calibration_R3.