mlsanityData to 7 Oct 2026

Where Bits Matter in World Model Planning: A Paired Mixed-Bit Study for Efficient Spatial Reasoning

arXiv 2602.11882 ↗First posted 12 Feb 2026cs.LG cs.AI cs.CVcited by 0

On the map: World models → World models for language agents

also in 1 more
Abstract (arXiv)

Efficient spatial reasoning requires world models that remain reliable under tight precision budgets. We study whether low-bit planning behavior is determined mostly by total bitwidth or by where bits are allocated across modules. Using DINO-WM on the Wall planning task, we run a paired-goal mixed-bit evaluation across uniform, mixed, asymmetric, and layerwise variants under two planner budgets. We observe a consistent three-regime pattern: 8-bit and 6-bit settings remain close to FP16, 3-bit settings collapse, and 4-bit settings are allocation-sensitive. In that transition region, preserving encoder precision improves planning relative to uniform quantization, and near-size asymmetric variants show the same encoder-side direction. In a later strict 22-cell replication with smaller per-cell episode count, the mixed-versus-uniform INT4 sign becomes budget-conditioned, which further highlights the sensitivity of this transition regime. These findings motivate module-aware, budget-aware quantization policies as a broader research direction for efficient spatial reasoning. Code and run artifacts are available at https://github.com/suraj-ranganath/DINO-MBQuant.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in World models for language agents, one of 11 directions in World models. Builds on 15 of its 15 references that are in the map. First posted on arXiv 12 Feb 2026.

Builds on

15 of 15 references.

References · 15

  1. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  2. 2022GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformerscited by 1,838
  3. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  4. 2018World Modelscited by 1,444
  5. 2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,338
  6. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
  7. 2018Learning Latent Dynamics for Planning from Pixelscited by 1,221
  8. 2023Mastering Diverse Domains through World Modelscited by 1,193
  9. 2022LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scalecited by 881
  10. 2018HAQ: Hardware-Aware Automated Quantization with Mixed Precisioncited by 426
  11. 2024DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planningcited by 357
  12. 2019HAWQ-V2: Hessian Aware trace-Weighted Quantization of Neural Networkscited by 250
3 more
  1. 2023FlatFormer: Flattened Window Attention for Efficient Point Cloud Transformercited by 72
  2. 2025Low-bit Model Quantization for Deep Neural Networks: A Surveycited by 11
  3. 2026An Empirical Study of World Model Quantizationcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Do LLMs Build Spatial World Models? Evidence from Grid-World Maze Tasks
  2. 2025From Building Blocks to Planning: Multi-Step Spatial Reasoning in LLMs with Reinforcement Learning
  3. 2024VSP: Assessing the dual challenges of perception and reasoning in spatial planning tasks for VLMs
  4. 2023Can Large Language Models be Good Path Planners? A Benchmark and Investigation on Spatial-temporal Reasoning
  5. 2026Rule-Compliant Visual Spatial Planning for Multimodal Large Language Models
  6. 2025SpatialBench: Benchmarking Multimodal Large Language Models for Spatial Cognition
  7. 2025Probing the effectiveness of World Models for Spatial Reasoning through Test-time Scaling
  8. 2026LAST: Leveraging Tools as Hints to Enhance Spatial Reasoning for Multimodal Large Language Models