RARE-Positive

This repository contains the final sample-level Positive checkpoints from RARE: Reinforcement Learning with Adaptive Rubric Evolution for Open-Ended Generation. Positive replacements express quality requirements that remain unmet by the current on-policy scripts.

Subfolder	Production style	Final RL step
2d_animation	2D animation	132
3d_stopmotion	3D and stop-motion animation	108
live_action	Live action	94

The held-out scores are 76.37, 75.59, and 69.91 for the three styles, respectively, for an equally weighted average of 73.96. This is a 1.98-point gain over the matched Sample-Static baseline.

Each subfolder is an independent inference-only full checkpoint derived from Qwen/Qwen3-4B-Instruct-2507; optimizer and trainer state are not included. The exact launcher and settings are documented in the RARE repository. Use the corresponding configuration from EurekaTian/RARE for the released system instruction and query format. This card does not add a separate convenience inference command that was not part of the released pipeline. The weights are licensed under Apache-2.0; the dataset is separately licensed under CC BY 4.0. The canonical paper citation will be added after publication metadata is finalized.