| Challenge: | Large vision–language models process images at native or high resolution to remain effective across tasks. |
| Approach: | They propose a lightweight preprocessing module that predicts the minimum sufficient input resolution for large vision–language models. |
| Outcome: | CARES predicts when a pre-trained VLM's response converges to its peak ability to answer correctly, reducing compute by up to 80%. |
Similar Papers
Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for estimating attention importance for tokens are ineffective . dLLMs require bidirectional attention, which limits inference efficiency . |
| Approach: | They propose a training-free attention sparsification framework for efficient long-context inference . they propose 'sink-aware pruning strategy' to accurately estimate and remove redundant computation . |
| Outcome: | The proposed approach offers 29 lossless speedup under 32K context length. |
Losing Visual Needles in Image Haystacks: Vision Language Models are Easily Distracted in Short and Long Contexts (2024.findings-emnlp)
Copied to clipboard
| Challenge: | evaluators of long-context vision language models (VLMs) have not kept up with the rapid development of open-weight long-constraint language models. |
| Approach: | They propose a dynamic benchmark generator for evaluating long-context reasoning in vision language models. |
| Outcome: | The proposed model can ignore irrelevant information when answering queries, showing that current models lack this capability. |
Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects (2026.findings-acl)
Copied to clipboard
Jun Zhang, Yicheng Ji, Feiyang Ren, Yihang Li, Bowen Zeng, Zonghao Chen, Ke Chen, Lidan Shou, Gang Chen, Huan Li
| Challenge: | Large Vision-Language Models are hindered by a systemic efficiency barrier known as visual token dominance. |
| Approach: | They propose a systematic taxonomy of efficiency techniques structured around the inference lifecycle . they examine visual encoding, prefilling, and decoding to understand bottlenecks . |
| Outcome: | The proposed techniques reveal how upstream decisions dictate downstream bottlenecks . the proposed techniques include hybrid compression and modality-aware decoding . |
OLIVE: Object Level In-Context Visual Embeddings (2024.acl-long)
Copied to clipboard
| Challenge: | Existing vision-language models lack fine-grained object-level understanding and grounding . existing models implicitly align text tokens with image patch tokens, which is ineffective for embedding alignment at the same granularity and introduces noisy spurious background features. |
| Approach: | They propose a method to prompt large language models with in-context visual object vectors . this method allows for controllable object-level reasoning . |
| Outcome: | The proposed method achieves competitive referring object classification and captioning performance while offering zero-shot generalization and robustness to visually challenging contexts. |
AdaV: Adaptive Text-visual Redirection for Vision-Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Vision-language models often generate excessive visual tokens, leading to poor performance . a novel training-free visual token pruning method is proposed to improve performance despite the computational cost associated with VLMs. |
| Approach: | They propose a training-free visual token pruning method that reduces biased token pruning . they plan to open-source the code upon publication . |
| Outcome: | The proposed method reduces biased token pruning and enhances model robustness with limited visual token budget. |
Efficient Architectures for High Resolution Vision-Language Models (2025.coling-main)
Copied to clipboard
| Challenge: | Recent advances in vision-Language Models (VLMs) have limited accuracy of fine details within high resolution images, which limits performance in multiple tasks. |
| Approach: | They propose a new architecture that efficiently processes high-resolution images while training fewer parameters than similarly sized VLMs. |
| Outcome: | The proposed architecture achieves high efficiency while maintaining strong performance in tasks that require fine-grained image understanding and/or handling of scene-text. |
Task-Aware Resolution Optimization for Visual Large Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing visual large language models pre-assume a fixed resolution for downstream tasks, leading to sub-optimal performance. |
| Approach: | They propose a formula to determine the optimal resolution for a given vision-language task . they then propose 'parameter-efficient' fine-tuning technique to extend the visual input resolution . |
| Outcome: | The proposed method is based on rigorous experiments on vision-language tasks. |
Look Less, Reason More: Rollout-Guided Adaptive Pixel-Space Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Recent work has shown promise by incorporating pixel-level visual information into the reasoning process, enabling VLMs to access high-resolution visual details during their thought process. |
| Approach: | They propose a framework that dynamically determines necessary pixel-level operations based on the input query. |
| Outcome: | The proposed model achieves 73.4% accuracy on HR-Bench 4K while maintaining a tool usage ratio of only 20.1%, improving accuracy and reducing tool usage by 66.5% compared to the previous methods. |
Mask Tokens as Prophet: Fine-Grained Cache Eviction for Efficient dLLM Inference (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing cache eviction strategies for autoregressive language models fail to account for the role of mask tokens and specific characteristics in dLLMs. |
| Approach: | They propose a training-free cache eviction framework tailored to dLLMs that denies a fully masked sequence and allows parallel decoding at the expense of memory and computation. |
| Outcome: | The proposed framework reduces the cost of memory and cache eviction and improves efficiency by reducing allocation in intermediate layers and concentrating resources on prompt-preferring heads. |
Efficient Sparse Attention needs Adaptive Token Release (2024.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide array of text-centric tasks, however, their ‘large’ scale introduces significant computational and storage challenges, particularly in managing the key-value states of the transformer, which limits their wider applicability. |
| Approach: | They propose to release resources from caches and rebuild key-value states by a lightweight controller module to approximate an ideal top-K sparse attention. |
| Outcome: | The proposed method achieves a significant throughput improvement of 221.8% over full attention and a model with 7 billion tokens. |