mlsanityData to 7 Oct 2026

BabyVision: Visual Reasoning Beyond Language

arXiv 2601.06521 ↗First posted 10 Jan 2026cs.CV cs.CLcited by 39

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

While humans develop core visual skills long before acquiring language, contemporary Multimodal LLMs (MLLMs) still rely heavily on linguistic priors to compensate for their fragile visual understanding. We uncovered a crucial fact: state-of-the-art MLLMs consistently fail on basic visual tasks that humans, even 3-year-olds, can solve effortlessly. To systematically investigate this gap, we introduce BabyVision, a benchmark designed to assess core visual abilities independent of linguistic knowledge for MLLMs. BabyVision spans a wide range of tasks, with 388 items divided into 22 subclasses across four key categories. Empirical results and human evaluation reveal that leading MLLMs perform significantly below human baselines. Gemini3-Pro-Preview scores 49.7, lagging behind 6-year-old humans and falling well behind the average adult score of 94.1. These results show despite excelling in knowledge-heavy evaluations, current MLLMs still lack fundamental visual primitives. Progress in BabyVision represents a step toward human-level visual perception and reasoning capabilities. We also explore solving visual reasoning with generation models by proposing BabyVision-Gen and automatic evaluation toolkit. Our code and benchmark data are released at https://github.com/UniPat-AI/BabyVision for reproduction.

Key facts. Cited by 39 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 14 of its 35 references that are in the map. First posted on arXiv 10 Jan 2026.

Builds on

14 of 35 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025MMGR: Multi-Modal Generative Reasoning Benchmark and Evaluation16 Dec
  2. 2025InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency25 Aug
  3. 2025MiMo-VL Technical Report4 Jun
  4. 2025Kimi-VL Technical Report10 Apr
  5. 2024MMEvalPro: Calibrating Multimodal Benchmarks Towards Trustworthy and Efficient Evaluation29 Jun
  6. 2024Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset22 Feb

Other references · 8

  1. 2025Qwen3-VL Technical Reportcited by 2,534
  2. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  3. 2023MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Modelscited by 1,624
  4. 2023Gemini: A Family of Highly Capable Multimodal Modelscited by 810
  5. 2024Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMscited by 719
  6. 2025Humanity's Last Examcited by 467
  7. 2025Step-3 is Large yet Affordable: Model-system Co-design for Cost-effective Decodingcited by 52
  8. 2025DrVD-Bench: Do Vision-Language Models Reason Like Human Doctors in Medical Image Diagnosis?cited by 10

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 39.

39 papers.

Sort:

By area:

  1. 2026Kimi K2.5: Visual Agentic Intelligencecited by 449
  2. 2026Kimi K3: Open Frontier Intelligencecited by 99
  3. 2026Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perceptioncited by 47
  4. 2026LongCat-Next: Lexicalizing Modalities as Discrete Tokenscited by 30
  5. 2026SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecturecited by 27
  6. 2026Step-wise Rubric Rewards for LLM Reasoningcited by 11
  7. 2026HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoningcited by 5
  8. 2026VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoningcited by 4
  9. 2026PerceptionBench: Evaluating Atomic Visual Perception in Multimodal Large Language Modelscited by 3
  10. 2026SenseNova-U1.5: Towards Native Unified Visual Intelligencecited by 3
  11. 2026Reasoning or Pattern Matching? Probing Large Vision-Language Models with Visual Puzzlescited by 2
  12. 2026Visual Reasoning Benchmark: Evaluating Multimodal LLMs on Classroom-Authentic Visual Problems from Primary Educationcited by 2
27 more
  1. 2026The Expense of Seeing: Attaining Trustworthy Multimodal Reasoning Within the Monolithic Paradigmcited by 2
  2. 2026Bridging Structure and Language: Graph-Based Visual Reasoning for Autonomous Road Understandingcited by 2
  3. 2026DeepLatent: Think with Images via Parallel Latent Visual Reasoningcited by 1
  4. 2026Beacon: Knowing When and How to Perform Agentic Visual Reasoningcited by 1
  5. 2026Hard Vision, Easy Vision: What GPT-6 Astra Reveals Across Computer Visioncited by 1
  6. 2026Semantic Richness or Geometric Reasoning? The Fragility of VLM's Visual Invariancecited by 0
  7. 2026EXAONE 4.5 Technical Reportcited by 0
  8. 2026MedVIGIL: Evaluating Trustworthy Medical VLMs Under Broken Visual Evidencecited by 0
  9. 2026The Cartesian Shortcut: Re-evaluate Vision Reasoning in Polar Coordinate Spacecited by 0
  10. 2026VLMs Trace Without Tracking: Diagnosing Failures in Visual Path Followingcited by 0
  11. 2026EgoBabyVLM: Benchmarking Cross-Modal Learning from Naturalistic Egocentric Video Datacited by 0
  12. 2026Perceive, Interact, Reason: Building Tool-Augmented Visual Agents for Spatial Reasoningcited by 0
  13. 2026Objects Before Words: Object-First Inductive Biases for Grounding Language in Child-View Videocited by 0
  14. 2026Blind-Spots-Bench: Evaluating Blind Spots in Multimodal Modelscited by 0
  15. 2026The Ebb and Flow of Multimodal Focus: Scheduling Visual Relay Windows for Grounded VLM Reasoningcited by 0
  16. 2026An Exam for Active Observerscited by 0
  17. 2026Spatial-IQ: Deconstructing Spatial Intelligence via Hierarchical Capability Testscited by 0
  18. 2026CoM3eT: A foundation model for medical image analysis through federated, multidimensional context integrationcited by 0
  19. 2026Projector Is All You Traincited by 0
  20. 2026Aphanta: Diagnosing Task-Aligned Image-Edited Intermediates for Multimodal Reasoningcited by 0
  21. 2026Thinking in Pictures: A Systematic Benchmark for Reasoning-driven Image Generationcited by 0
  22. 2026GenPuzzle: Benchmarking Visual Reasoning in Image Generation Modelscited by 0
  23. 2026MindTopo: Can Foundation Models Reason in Topological Space?cited by 0
  24. 2026SolveEdit: Benchmarking Visual Problem Solving in Generative Modelscited by 0
  25. 2026EpiCon: Collective Agent Learning through Co-Evolving Multimodal Memorycited by 0
  26. 2026Where MLLMs Fail and Why: Causal Task Decomposition for Capability Failure Diagnosiscited by 0
  27. 2026VISTA: A Visual Harness for Reasoning in an Interactive Worldcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026KidVis: Do Multimodal Large Language Models Possess the Visual Perceptual Capabilities of a 6-Year-Old?
  2. 2025BabyVLM-V2: Toward Developmentally Grounded Pretraining and Benchmarking of Vision Foundation Models
  3. 2025BabyVLM: Data-Efficient Pretraining of VLMs Inspired by Infant Learning
  4. 2025NanoVLMs: How small can we go and still make coherent Vision Language Models?
  5. 2024What is the Visual Cognition Gap between Humans and Multimodal LLMs?
  6. 2025Do You See Me : A Multidimensional Benchmark for Evaluating Visual Perception in Multimodal LLMs
  7. 2026Mind's Eye: A Benchmark of Visual Abstraction, Transformation and Composition for Multimodal LLMs
  8. 2025The Percept-V Challenge: Can Multimodal LLMs Crack Simple Perception Problems?