Architectural Sampling: Test-Time Scaling via Computational Diversity in Frozen Vision-Language Models
cs.CV, cs.AI
Submitted: 2026-10-01
Updated: 2026-10-01
Project page: https://akshit21112002.github.io/Architectural-Sampling/ABSTRACT
License: http://creativecommons.org/licenses/by/4.0/
The gist: Test-time scaling often seeks better answers by sampling multiple responses from a frozen model, yet conventional temperature sampling generates every candidate along the same fixed computation path.
Terminology
Abstract
Test-time scaling often seeks better answers by sampling multiple responses from a frozen model, yet conventional temperature sampling generates every candidate along the same fixed computation path. We introduce architectural sampling, a training-free method that generates candidates through distinct forward computations by reusing selected blocks of decoder layers. Varying the block location and repetition count introduces computational diversity without updating model weights or adding auxiliary parameters. Across five Qwen checkpoints and twelve multimodal benchmarks, architectural sampling improves pass@9 over standard-path temperature sampling by 6.58 percentage points on average at the same nine-candidate budget. Reusing early layers yields the strongest gains, and the improvement in candidate coverage persists even under greedy decoding. The resulting candidates show lower lexical overlap and improve accuracy when used as rollouts for label-free test-time reinforcement learning. These findings extend the benefits of our architectural sampling beyond candidate coverage, demonstrating more effective learning from a model's own outputs.
Sources
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- Training-Free Looped Transformers
- Evaluating Large Language Models Trained on Code
- BLINK: Multimodal Large Language Models Can See but Not Perceive
- Skip a Layer or Loop it? Test-Time Depth Adaptation of Pretrained LLMs
- Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering
- MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts
- Inner Loop Inference for Pretrained Transformers: Unlocking Latent Capabilities Without Training
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- CountQA: How Well Do MLLMs Count in the Wild?
- Gemma 3 Technical Report
- Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs
- Looping Back to Move Forward: Recursive Transformers for Efficient and Flexible Large Multimodal Models
- MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI
- InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models