mlsanityData to 7 Oct 2026

Can Vision Replace Text in Working Memory? Evidence from Spatial n-Back in Vision-Language Models

arXiv 2602.04355 ↗First posted 4 Feb 2026cs.CLcited by 0

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

Abstract (arXiv)

Working memory is a central component of intelligent behavior, providing a dynamic workspace for maintaining and updating task-relevant information. Recent work has used n-back tasks to probe working-memory-like behavior in large language models, but it is unclear whether the same probe elicits comparable computations when information is carried in a visual rather than textual code in vision-language models. We evaluate Qwen2.5 and Qwen2.5-VL on a controlled spatial n-back task presented as matched text-rendered or image-rendered grids. Across conditions, models show reliably higher accuracy and d' with text than with vision. To interpret these differences at the process level, we use trial-wise log-probability evidence and find that nominal 2/3-back often fails to reflect the instructed lag and instead aligns with a recency-locked comparison. We further show that grid size alters recent-repeat structure in the stimulus stream, thereby changing interference and error patterns. These results motivate computation-sensitive interpretations of multimodal working memory.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 8 of its 46 references that are in the map. First posted on arXiv 4 Feb 2026.

Builds on

8 of 46 references.

References · 8

  1. 2025Qwen2.5-VL Technical Reportcited by 5,418
  2. 2024Qwen2.5 Technical Reportcited by 4,713
  3. 2024GPT-4o System Cardcited by 4,467
  4. 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,175
  5. 2025Qwen2.5-Omni Technical Reportcited by 704
  6. 2025DeepSeek-OCR: Contexts Optical Compressioncited by 174
  7. 2026AgentOCR: Reimagining Agent History via Optical Self-Compressioncited by 22
  8. 2025Towards Long-window Anchoring in Vision-Language Model Distillationcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Words or Vision: Do Vision-Language Models Have Blind Faith in Text?
  2. 2025Position: Reasoning After Perception Means Reasoning Without Vision
  3. 2026Seeing Beyond Redundancy: Task Complexity's Role in Vision Token Specialization in VLLMs
  4. 2026Knowing Isn't Always Saying: When Do Spatial Encodings Reach Answers in Vision-Language Models?
  5. 2025Visible Yet Unreadable: A Systematic Blind Spot of Vision Language Models Across Writing Systems
  6. 2026Thinking With Tools, Not With Pixels: Tool Calls as Text Scaffolds for Visual Reasoning
  7. 2026Reasoning Dynamics and the Limits of Monitoring Modality Reliance in Vision-Language Models
  8. 2026Understanding Cross-Modal Contributions in Continual Vision-Language Models: A Theoretical Perspective