mlsanityData to 14 Oct 2026

LOCI: A Locator-Critic with Refinement Loop

arXiv 2608.30959 ↗First posted 31 Aug 2026cs.CV cs.AIcited by 0

On the map: Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning

Abstract (arXiv)

Vision-Language Models (VLMs) still struggle on tasks requiring complex visual understanding. We argue that the core issue is not high-level reasoning, but instead failing to locate critical details in the image. Due to this shortcoming, VLMs generate often plausible but incorrect reasoning based on flawed perceptual grounding. To address this, we propose Locator-Critic (LOCI), a training-free framework that decouples visual search from evidence verification. LOCI employs a Locator agent to propose candidate visual evidence and a separate Critic agent to evaluate its relevance and sufficiency. These agents engage in an iterative refinement loop, progressively improving the evidence until it is adequate to answer the given question. This decoupled, self-correcting process yields substantial performance gains, achieving state-of-the-art results on multiple complex visual benchmarks. LOCI improves accuracy for both open-weight models like Qwen3-VL (+12.1 on V*, +5.8 on HR-Bench and +11.2 on VisualProbe-Hard) and proprietary models like Gemini 2.5 Pro (+8.9 on V*, +4.3 on HR-Bench, +4.8 on VisualProbe-Hard).

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 23 of its 37 references that are in the map. First posted on arXiv 31 Aug 2026.

Builds on

23 of 37 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Generative Universal Verifier as Multimodal Meta-Reasoner15 Oct
  2. 2025Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search9 Sep
  3. 2025MMC: Iterative Refinement of VLM Reasoning via MCTS-based Multimodal Critique15 Apr
  4. 2023Griffon: Spelling out All Object Locations at Any Granularity with Large Language Models24 Nov

Other references · 19

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  2. 2024GPT-4o System Cardcited by 4,588
  3. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,796
  4. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,442
  5. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,719
  6. 2025OpenAI GPT-5 System Cardcited by 989
  7. 2023MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Actioncited by 517
  8. 2022Visual Programming: Compositional visual reasoning without trainingcited by 516
  9. 2023V*: Guided Visual Search as a Core Mechanism in Multimodal LLMscited by 482
  10. 2025DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learningcited by 410
  11. 2023Scaling Open-Vocabulary Object Detectioncited by 343
  12. 2025Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learningcited by 278
7 more
  1. 2025ToRL: Scaling Tool-Integrated RLcited by 169
  2. 2025Thyme: Think Beyond Imagescited by 167
  3. 2024Divide, Conquer and Combine: A Training-Free Framework for High-Resolution Image Perception in Multimodal Large Language Modelscited by 161
  4. 2024Generating CAD Code with Vision-Language Models for 3D Designscited by 62
  5. 2024Critic-V: VLM Critics Help Catch VLM Errors in Multimodal Reasoningcited by 50
  6. 2025DyFo: A Training-Free Dynamic Focus Visual Search for Enhancing LMMs in Fine-Grained Visual Understandingcited by 49
  7. 2025LLaVA-Critic-R1: Your Critic Model is Secretly a Strong Policy Modelcited by 35

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Improving Visual Reasoning with Iterative Evidence Refinement
  2. 2025Draft and Refine with Visual Experts
  3. 2026DeepScan: A Training-Free Framework for Visually Grounded Reasoning in Large Vision-Language Models
  4. 2025VERIFY: A Benchmark of Visual Explanation and Reasoning for Investigating Multimodal Reasoning Fidelity
  5. 2025VisRes Bench: On Evaluating the Visual Reasoning Capabilities of VLMs
  6. 2024UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling
  7. 2026Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?
  8. 2025Diagnosing Visual Reasoning: Challenges, Insights, and a Path Forward