AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Discriminative Visual Process Rewards for Scaling Thinking at Test-Time with Images

Bo-Wen Yin, Qize Yang, Boyuan Sun, Xihan Wei, Qibin Hou

ICML 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a discriminative process reward model that scores intermediate visual steps (cropping, annotation, localization, sketches) in thinking-with-images reasoning, enabling test-time scaling for multimodal LLMs.

Problem

Existing multimodal process reward models evaluate only textual reasoning and cannot judge the correctness of intermediate visual steps in "thinking with images" reasoning, leaving a gap for verifying and scaling visual reasoning at test time.

Approach

A discriminative visual process reward model that judges the correctness of intermediate visual actions (e.g., cropping, annotation, localization, sketching) rather than only textual reasoning steps, used to guide test-time scaling of thinking with images.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use DiscPRM instead of VisualPRM for Best-of-8 selectionDeepeyes · Perception benchmarksaverage accuracy: 71.5 -> 75.5Improved
use DiscPRM instead of VisualPRM for Best-of-8 selectionThyme-7B · Reasoning benchmarksaverage accuracy: 43.8 -> 47.4Improved
use DiscPRM instead of VisualPRM for Best-of-8 selectionThyme-7B · General benchmarksaverage accuracy: 72.6 -> 75.9Improved
use DiscPRM rather than Self-Consistency majority votingDeepeyes · MathVisionaccuracy: 27.3 -> 36.4Improved
use Self-Consistency majority voting over 8 samplesThyme-7B · V*accuracy: 82.2 -> 82.8No clear effect
use value-based instead of advantage-based formulation (ablation)DiscPRM · MathVistaaccuracy: 72.4 -> 75.5Improved
supervise all steps instead of stopping at first incorrect step (ablation)DiscPRM · MathVistaaccuracy: 73.5 -> 75.5Improved
supervise all steps instead of stopping at first incorrect step (ablation)DiscPRM · VABenchmacro F1: 63.4 -> 65.9Improved
average step scores instead of max pooling (ablation)DiscPRM · MathVistaaccuracy: 73.1 -> 75.5Improved
average step scores instead of min pooling (ablation)DiscPRM · MathVistaaccuracy: 73.5 -> 75.5Improved
raise data sampling temperature from 0.7 to 1.3 (ablation)DiscPRM · V*accuracy: 89.5 -> 88.3Hurt
lower data sampling temperature from 0.7 to 0.4 (ablation)DiscPRM · V*accuracy: 89.5 -> 88.8No clear effect

Experimental setup

Builds on: process reward models thinking with images

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy