# Diversity Matters: Revisiting Test-Time Compute in Vision-Language Models

arXiv: https://arxiv.org/abs/2605.30713 · HTML: https://mlsanity.com/paper/2605.30713/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 17 of its 46 references that are in the map.
- First posted on arXiv 29 May 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [Prompt learning and adaptation of CLIP](https://mlsanity.com/mllm/prompt-learning-and-adaptation-of-clip/)

## Builds on (most cited first, 17 of 17)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [Demystifying Long Chain-of-Thought Reasoning in LLMs](https://mlsanity.com/paper/2502.03373/) (2025)
- [Pixtral 12B](https://mlsanity.com/paper/2410.07073/) (2024)
- [MME-CoT: Benchmarking Chain-of-Thought in Large Multimodal Models for Reasoning Quality, Robustness, and Efficiency](https://mlsanity.com/paper/2502.09621/) (2025)
- [Efficient Test-Time Scaling via Self-Calibration](https://mlsanity.com/paper/2503.00031/) (2025)
- [VGR: Visual Grounded Reasoning](https://mlsanity.com/paper/2506.11991/) (2025)
- [UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling](https://mlsanity.com/paper/2408.04810/) (2024)
- [Look, Remember and Reason: Grounded reasoning in videos with language models](https://mlsanity.com/paper/2306.17778/) (2023)
- [VISTA: Enhancing Vision-Text Alignment in MLLMs via Cross-Modal Mutual Information Maximization](https://mlsanity.com/paper/2505.10917/) (2025)
- [Temporalizing Confidence: Evaluation of Chain-of-Thought Reasoning with Signal Temporal Logic](https://mlsanity.com/paper/2506.08243/) (2025)
- [Style over Substance: Distilled Language Models Reason Via Stylistic Replication](https://mlsanity.com/paper/2504.01738/) (2025)
- [Test-Time Consistency in Vision Language Models](https://mlsanity.com/paper/2506.22395/) (2025)

## Built on by (most cited first, 1 of 1)

- [It's the Decoding Format, Not the Perturbation: Auditing Consistency-Based Selection for Vision-Language Test-Time Scaling](https://mlsanity.com/paper/2608.01207/) (2026)

## Abstract (arXiv)

Test-time compute (TTC) strategies have emerged as a lightweight approach to boost reasoning in large language models (LLMs). However, their application and benefits for vision-language models (VLMs) remain underexplored. We present a systematic study of TTC across seven VLMs and six benchmarks, specifically analyzing feature-based scoring and majority voting methods. We find that feature heuristics fail and voting yields only modest gains in single-model settings. We theoretically show that this limitation stems from a lack of prediction diversity: when outputs are highly correlated, voting provides little benefit. In contrast, multi-model ensembles offer richer diversity, yet standard majority voting fails to account for varying model capabilities. To address this, we propose Entropy-based TTC (ETTC), which selects the most confident prediction based on predictive entropy. Our method reduces to majority voting in the single-model case, but in model ensembles, it leverages confidence disparities to prioritize stronger models. We prove that ETTC outperforms majority voting under mild assumptions and empirically demonstrate that it consistently surpasses both voting and the best individual model. Crucially, our results show that smaller models can synergistically enhance larger ones, unlocking ensembling gains not achievable with standard strategies.