You Only Judge Once: Multi-response Reward Modeling in a Single Forward Pass

Dimensions overall score 8.0

• A multimodal reward model that scores all candidate responses in a single forward pass, achieving significant speedups and outperforming existing models for improved generation quality. Conventional discriminative reward models evaluate each response independently, requiring multiple forward passes, one for each potential response.problem 35%
  Evidencepartial

  A multimodal reward model that scores all candidate responses in a single forward pass, achieving significant speedups and outperforming existing models for improved generation quality. Conventional discriminative reward models evaluate each response independently, requiring multiple forward passes, one for each potential response.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

• We present a discriminative multimodal reward model that scores all candidate responses in a single forward pass. Conventional discriminative reward models evaluate each response independently, requiring multiple forward passes, one for each potential response.method 35%
  Evidencepartial

  We present a discriminative multimodal reward model that scores all candidate responses in a single forward pass. Conventional discriminative reward models evaluate each response independently, requiring multiple forward passes, one for each potential response.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

• ScienceToStartup currently rates this 8.0/10 on the public viability pass. To enable $N$-way reward evaluation beyond existing pairwise benchmarks, we construct two new benchmarks: (1) MR$^2$Bench-Image contains human-annotated rankings over responses from 8 diverse models; (2) MR$^2$Bench-Video is a large-scale video-based reward benchmark derived from 94K crowdsourced pairwise human judgments over video question-answering spanning 19 models, denoised via preference graph ensemble. Code availability is flagged in the production record; the public repository link still needs proof alignment.result 35%
  Evidencepartial

  ScienceToStartup currently rates this 8.0/10 on the public viability pass. To enable $N$-way reward evaluation beyond existing pairwise benchmarks, we construct two new benchmarks: (1) MR$^2$Bench-Image contains human-annotated rankings over responses from 8 diverse models; (2) MR$^2$Bench-Video is a large-scale video-based reward benchmark derived from 94K crowdsourced pairwise human judgments over video question-answering spanning 19 models, denoised via preference graph ensemble. Code availability is flagged in the production record; the public repository link still needs proof alignment.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

• Multimodal Reward Modeling moved forward this cycle; last verified April 2026. Public score 8.0/10. Production flags indicate code availability.why now 35%
  Evidencepartial

  Multimodal Reward Modeling moved forward this cycle; last verified April 2026. Public score 8.0/10. Production flags indicate code availability.

  Implicationpartial

  Abstract-backed fallback claim; anchored extraction has not materialized a public claim row yet.

  Verificationpartial

  partial

Competitive landscape

A multimodal reward model that scores all candidate responses in a single forward pass, achieving significant speedups and outperforming existing models for improved generation quality.