Metraipolis › archive › dc_697461e59333
CARE: Contrastive Anchor-based Rubric Evolution for Large Language Model Post-Training — arXiv:2609.00892
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.00892
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T06:32:16.701Z
- SHA-256 of the archived text
2dab8d27ac5df1fba2ce0556fb4a5d9fd3ecc18d84249a5b1ab659bbaea32443- Stable URL of this page
https://metraipolis.com/documento/dc_697461e59333
How to cite
CARE: Contrastive Anchor-based Rubric Evolution for Large Language Model Post-Training — arXiv:2609.00892 — arXiv (en), https://arxiv.org/abs/2609.00892 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_697461e59333 (Metraipolis archive dc_697461e59333, sha256 2dab8d27ac5df1fba2ce0556fb4a5d9fd3ecc18d84249a5b1ab659bbaea32443)
BibTeX
@misc{dc697461e59333,
title = {CARE: Contrastive Anchor-based Rubric Evolution for Large Language Model Post-Training — arXiv:2609.00892},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.00892},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_697461e59333; sha256 2dab8d27ac5df1fba2ce0556fb4a5d9fd3ecc18d84249a5b1ab659bbaea32443},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00892. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Rubric-based reinforcement learning decomposes open-ended instructions into prompt-specific, flexible rubrics, making it better suited than reinforcement learning with verifiable rewards for post-training LLMs on open-ended tasks. However, static rubrics are inevitably hacked as the policy evolves, and existing dynamic approaches introduce new problems: undirected rubric extraction, unreliable hack detection, and unbounded rubric proliferation. We propose $\textbf{CARE}$ ($\textbf{C}$ontrastive $\textbf{A}$nchor-based $\textbf{R}$ubric $\textbf{E}$volution), which grounds every rubric evolution step in a high-quality anchor response generated by a frontier model conditioned on the prompt and its rubrics. At each training step, CARE contrasts the highest-scoring rollout against the anchor, enabling two complementary mechanisms: an Adaptive branch that reactively repairs reward misspecification; and a Chase branch that proactively converts frontier-level quality gaps into sharper rubrics. Together, the two branches $\textbf{maintain discriminative accuracy in the high-reward region}$---the precise region where reward over-optimization mostly originates. Experiments on WildChecklist-9K with Qwen2.5-7B-Base and Qwen2.5-7B-Instruct show that CARE achieves state-of-the-art performance on Arena-Hard-2.0, InfoBench, and FollowBench, and is the $\textbf{only}$ method whose win rate against GPT-4.1 anchor responses shows sustained improvement throughout 300 training steps; additional results on Llama-3.1-8B-Instruct and Qwen3-8B further indicate that CARE generalizes across model families.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.00892 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T06:15:12.659Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).