# FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering

arXiv: https://arxiv.org/abs/2506.21710 · HTML: https://mlsanity.com/paper/2506.21710/

## Key facts

- Cited by 20 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 38 of its 46 references that are in the map.
- First posted on arXiv 26 Jun 2025.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 30 of 38)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [VQA: Visual Question Answering](https://mlsanity.com/paper/1505.00468/) (2015)
- [Rich feature hierarchies for accurate object detection and semantic segmentation](https://mlsanity.com/paper/1311.2524/) (2013)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [MiniCPM-V: A GPT-4V Level MLLM on Your Phone](https://mlsanity.com/paper/2408.01800/) (2024)
- [How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites](https://mlsanity.com/paper/2404.16821/) (2024)
- [Vision Transformers Need Registers](https://mlsanity.com/paper/2309.16588/) (2023)
- [Quantifying Attention Flow in Transformers](https://mlsanity.com/paper/2005.00928/) (2020)
- [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://mlsanity.com/paper/2401.06209/) (2024)
- [A-OKVQA: A Benchmark for Visual Question Answering using World Knowledge](https://mlsanity.com/paper/2206.01718/) (2022)
- [SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities](https://mlsanity.com/paper/2401.12168/) (2024)
- [An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models](https://mlsanity.com/paper/2403.06764/) (2024)
- [LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models](https://mlsanity.com/paper/2407.07895/) (2024)
- [FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPU](https://mlsanity.com/paper/2303.06865/) (2023)
- [Efficiently Scaling Transformer Inference](https://mlsanity.com/paper/2211.05102/) (2022)
- [V*: Guided Visual Search as a Core Mechanism in Multimodal LLMs](https://mlsanity.com/paper/2312.14135/) (2023)
- [PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction](https://mlsanity.com/paper/2410.17247/) (2024)
- [MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?](https://mlsanity.com/paper/2408.13257/) (2024)
- [LLaVA-UHD: an LMM Perceiving Any Aspect Ratio and High-Resolution Images](https://mlsanity.com/paper/2403.11703/) (2024)

## Built on by (most cited first, 20 of 20)

- [Q-Zoom: Query-Aware Adaptive Perception for Efficient Multimodal Large Language Models](https://mlsanity.com/paper/2604.06912/) (2026)
- [TangramPuzzle: Evaluating Multimodal Large Language Models with Compositional Spatial Reasoning](https://mlsanity.com/paper/2601.16520/) (2026)
- [CARE: Towards Clinical Accountability in Multi-Modal Medical Reasoning with an Evidence-Grounded Agentic Framework](https://mlsanity.com/paper/2603.01607/) (2026)
- [ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison](https://mlsanity.com/paper/2605.20278/) (2026)
- [PixelEyes: Decoupling Perception and Reasoning for Pinpoint Visual Evidence Seeking](https://mlsanity.com/paper/2607.00115/) (2026)
- [Unexplored flaws in multiple-choice VQA make benchmarking unreliable](https://mlsanity.com/paper/2511.22341/) (2025)
- [MRD: Multi-resolution Retrieval-Detection Fusion for High-Resolution Image Understanding](https://mlsanity.com/paper/2512.02906/) (2025)
- [Peek-a-Boo Reasoning: Contrastive Region Masking in MLLMs](https://mlsanity.com/paper/2512.08976/) (2025)
- [ActiveScope: Actively Seeking and Correcting Perception for MLLMs](https://mlsanity.com/paper/2606.24292/) (2026)
- [Region-Level Policy Optimization for Fine-grained MLLM Perception](https://mlsanity.com/paper/2609.19745/) (2026)
- [Efficient Cross-Country Data Acquisition Strategy for ADAS via Street-View Imagery](https://mlsanity.com/paper/2602.01836/) (2026)
- [ConFoThinking: Consolidated Focused Attention Driven Thinking for Visual Question Answering](https://mlsanity.com/paper/2603.00165/) (2026)
- [BoxTuning: Directly Injecting the Object Box for Multimodal Model Fine-Tuning](https://mlsanity.com/paper/2604.11136/) (2026)
- [BioELX: Context-Aware Cross-lingual Biomedical Entity Linking without Task-Specific Supervision](https://mlsanity.com/paper/2605.27380/) (2026)
- [ROVER: Routing Object-Centric Visual Evidence for Grounded Multi-Image Reasoning](https://mlsanity.com/paper/2605.27959/) (2026)
- [BVS: Bayesian Visual Search with Multimodal Large Language Model for Fine-grained Perception](https://mlsanity.com/paper/2607.03184/) (2026)
- [ChronoVision: Temporal Reasoning via Latent State Reconstruction](https://mlsanity.com/paper/2608.05631/) (2026)
- [VisLens: Single-Pass Interpretable Visual Search for Multimodal LLMs](https://mlsanity.com/paper/2608.30705/) (2026)
- [Pay More Attention To Text In High-Resolution MLLMs](https://mlsanity.com/paper/2609.23495/) (2026)
- [EviViT: Evidence-Adaptive Vision Transformers for Fine-Grained Perception](https://mlsanity.com/paper/2609.37123/) (2026)

## Abstract (arXiv)

While Multimodal Large Language Models (MLLMs) offer strong perception and reasoning capabilities for image-text input, Visual Question Answering (VQA) focusing on small image details still remains a challenge. Although visual cropping techniques seem promising, recent approaches have several limitations: the need for task-specific fine-tuning, low efficiency due to uninformed exhaustive search, or incompatibility with efficient attention implementations. We address these shortcomings by proposing a training-free visual cropping method, dubbed FOCUS, that leverages MLLM-internal representations to guide the search for the most relevant image region. This is accomplished in four steps: first, we identify the target object(s) in the VQA prompt; second, we compute an object relevance map using the key-value (KV) cache; third, we propose and rank relevant image regions based on the map; and finally, we perform the fine-grained VQA task using the top-ranked region. As a result of this informed search strategy, FOCUS achieves strong performance across four fine-grained VQA datasets and three types of MLLMs. It outperforms three popular visual cropping methods in both accuracy and efficiency, and matches the best-performing baseline, ZoomEye, while requiring 3 - 6.5 x less compute.