Pith. sign in

REVIEW 8 cited by

Multi-Step Visual Reasoning with Visual Tokens Scaling and Verification

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2506.07235 v1 pith:SVSH7KJF submitted 2025-06-08 cs.CV cs.CL

Tianyi Bai , Zengjie Hu , Fupeng Sun , Jiantao Qiu , Yizhen Jiang , Guangxin He , Bohan Zeng , Conghui He
show 2 more authors
Binhang Yuan Wentao Zhang
This is my paper
classification cs.CVcs.CL
keywords visualreasoningmllmsinferenceactionsdynamicenablinglanguage
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Multi-modal large language models (MLLMs) have achieved remarkable capabilities by integrating visual perception with language understanding, enabling applications such as image-grounded dialogue, visual question answering, and scientific analysis. However, most MLLMs adopt a static inference paradigm, encoding the entire image into fixed visual tokens upfront, which limits their ability to iteratively refine understanding or adapt to context during inference. This contrasts sharply with human perception, which is dynamic, selective, and feedback-driven. In this work, we introduce a novel framework for inference-time visual token scaling that enables MLLMs to perform iterative, verifier-guided reasoning over visual content. We formulate the problem as a Markov Decision Process, involving a reasoner that proposes visual actions and a verifier, which is trained via multi-step Direct Preference Optimization (DPO), that evaluates these actions and determines when reasoning should terminate. To support this, we present a new dataset, VTS, comprising supervised reasoning trajectories (VTS-SFT) and preference-labeled reasoning comparisons (VTS-DPO). Our method significantly outperforms existing approaches across diverse visual reasoning benchmarks, offering not only improved accuracy but also more interpretable and grounded reasoning processes. These results demonstrate the promise of dynamic inference mechanisms for enabling fine-grained, context-aware visual reasoning in next-generation MLLMs.

Discussion (0). Sign in to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. VABench: A Comprehensive Benchmark for Audio-Video Generation

    cs.CV 2025-12 unverdicted novelty 7.0 of 10

    VABench is a new multi-dimensional benchmark for evaluating synchronous audio-video generation across text-to-AV, image-to-AV, and stereo tasks.

  2. BVS: Bayesian Visual Search with Multimodal Large Language Model for Fine-grained Perception

    cs.CV 2026-07 conditional novelty 6.0 of 10

    BVS combines early-stop attention rollout priors with a scale-aware non-stationary kernel and GP-UCB to locate tiny objects in UHR images more accurately and with fewer MLLM queries than prior visual-search methods.

  3. ROVER: Routing Object-Centric Visual Evidence for Grounded Multi-Image Reasoning

    cs.CV 2026-05 unverdicted novelty 6.0 of 10

    ROVER introduces a learnable routing plugin for object-centric visual evidence in MLLMs via token triplets and differential attention, reporting gains on MM-GCoT and VideoEspresso when integrated into Qwen2.5-VL-7B.

  4. Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensembles

    cs.LG 2026-05 unverdicted novelty 6.0 of 10

    Maestro uses outcome-based RL to train a lightweight policy that orchestrates ensembles of frozen expert models and skills, reporting 70.1% average accuracy across ten multimodal benchmarks and outperforming GPT-5 and...

  5. Test-time Scaling over Perception: Resolving the Grounding Paradox in Thinking with Images

    cs.CV 2026-04 conditional novelty 6.0 of 10

    TTSP samples and filters multiple zoom-in exploration traces and iteratively consolidates validated observations into an Evidence Ledger, improving fine-grained multimodal reasoning on V* Bench, HR-Bench, TreeBench, a...

  6. Test-time Scaling over Perception: Resolving the Grounding Paradox in Thinking with Images

    cs.CV 2026-04 unverdicted novelty 5.0 of 10

    TTSP resolves the Grounding Paradox by treating perception as a scalable test-time process that generates, filters, and iteratively refines multiple visual exploration traces, outperforming baselines on high-resolutio...

  7. OpenWorldLib: A Unified Codebase and Definition of Advanced World Models

    cs.CV 2026-04 conditional novelty 4.0 of 10

    OpenWorldLib defines world models as perception-centered systems with interaction and long-term memory, and provides a modular inference codebase unifying interactive video, 3D, reasoning, and VLA tasks.

  8. OpenWorldLib: A Unified Codebase and Definition of Advanced World Models

    cs.CV 2026-04 unverdicted novelty 4.0 of 10

    OpenWorldLib offers a standardized codebase and definition for world models that combine perception, interaction, and memory to understand and predict the world.

Pith tools