Skip to content
AI News HubLIVE
Original source2 min read

Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria

Summary

This paper proposes Auto-Rubric as Reward (ARR), a framework that transforms reward modeling for multimodal generative models from implicit weight optimization to explicit criteria-based decomposition. ARR externalizes a VLM's internalized preference knowledge as prompt-specific rubrics, reducing evaluation biases. Combined with Rubric Policy Optimization (RPO), it outperforms pairwise reward models and VLM judges on text-to-image and image editing benchmarks. The work reveals that the alignment bottleneck is the lack of a factorized interface, not a deficit of knowledge.

SourcearXiv AIAuthor: Juanxi Tian, Fengyuan Liu, Jiaming Han, Yilei Jiang, Yongliang Wu, Yesheng Liu, Haodong Li, Furong Xu, Wanhua Li
Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[2605.08354] Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria

[Submitted on 8 May 2026]

Title:Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria

View a PDF of the paper titled Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria, by Juanxi Tian and 7 other authors

View PDF HTML (experimental)

Abstract:Aligning multimodal generative models with human preferences demands reward signals that respect the compositional, multi-dimensional structure of human judgment. Prevailing RLHF approaches reduce this structure to scalar or pairwise labels, collapsing nuanced preferences into opaque parametric proxies and exposing vulnerabilities to reward hacking. While recent Rubrics-as-Reward (RaR) methods attempt to recover this structure through explicit criteria, generating rubrics that are simultaneously reliable, scalable, and data-efficient remains an open problem. We introduce Auto-Rubric as Reward (ARR), a framework that reframes reward modeling from implicit weight optimization to explicit, criteria-based decomposition. Before any pairwise comparison, ARR externalizes a VLM's internalized preference knowledge as prompt-specific rubrics, translating holistic intent into independently verifiable quality dimensions. This conversion of implicit preference structure into inspectable, interpretable constraints substantially suppresses evaluation biases including positional bias, enabling both zero-shot deployment and few-shot conditioning on minimal supervision. To extend these gains into generative training, we propose Rubric Policy Optimization (RPO), which distills ARR's structured multi-dimensional evaluation into a robust binary reward, replacing opaque scalar regression with rubric-conditioned preference decisions that stabilize policy gradients. On text-to-image generation and image editing benchmarks, ARR-RPO outperforms pairwise reward models and VLM judges, demonstrating that explicitly externalizing implicit preference knowledge into structured rubrics achieves more reliable, data-efficient multimodal alignment, revealing that the bottleneck is the absence of a factorized interface, not a deficit of knowledge.

Comments: 28 pages, 10 figures, 11 tables

Subjects:

Artificial Intelligence (cs.AI)

Cite as: arXiv:2605.08354 [cs.AI]

(or arXiv:2605.08354v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2605.08354

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Juanxi Tian [view email] [v1] Fri, 8 May 2026 18:05:27 UTC (4,413 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria, by Juanxi Tian and 7 other authors

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.AI

new | recent | 2026-05

Change to browse by:

cs

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • ARR externalizes implicit preferences into explicit, interpretable rubrics, enabling zero-shot and few-shot deployment.
  • RPO distills multi-dimensional evaluation into a robust binary reward for stable policy gradients.
  • ARR-RPO achieves state-of-the-art performance on text-to-image and image editing tasks.
  • The bottleneck in multimodal alignment is the absence of a factorized interface, not knowledge deficit.

Highlights and analysis are generated automatically and may contain errors. Check the original source.

More recent stories on this topic · Models