# Direct Visual Grounding by Directing Attention of Visual Tokens

arXiv: https://arxiv.org/abs/2511.12738 · HTML: https://mlsanity.com/paper/2511.12738/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 50 of its 60 references that are in the map.
- First posted on arXiv 16 Nov 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)

## Builds on (most cited first, 30 of 50)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks](https://mlsanity.com/paper/1908.02265/) (2019)
- [ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning](https://mlsanity.com/paper/2203.10244/) (2022)
- [CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning](https://mlsanity.com/paper/1612.06890/) (2016)
- [CoCa: Contrastive Captioners are Image-Text Foundation Models](https://mlsanity.com/paper/2205.01917/) (2022)
- [Kosmos-2: Grounding Multimodal Large Language Models to the World](https://mlsanity.com/paper/2306.14824/) (2023)
- [Modeling Context in Referring Expressions](https://mlsanity.com/paper/1608.00272/) (2016)
- [Vision Transformers Need Registers](https://mlsanity.com/paper/2309.16588/) (2023)
- [Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs](https://mlsanity.com/paper/2406.16860/) (2024)
- [LISA: Reasoning Segmentation via Large Language Model](https://mlsanity.com/paper/2308.00692/) (2023)
- [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://mlsanity.com/paper/2401.06209/) (2024)
- [DeepSeek-VL: Towards Real-World Vision-Language Understanding](https://mlsanity.com/paper/2403.05525/) (2024)
- [SimVLM: Simple Visual Language Model Pretraining with Weak Supervision](https://mlsanity.com/paper/2108.10904/) (2021)
- [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)
- [Attention is not not Explanation](https://mlsanity.com/paper/1908.04626/) (2019)
- [GLaMM: Pixel Grounding Large Multimodal Model](https://mlsanity.com/paper/2311.03356/) (2023)
- [Pix2seq: A Language Modeling Framework for Object Detection](https://mlsanity.com/paper/2109.10852/) (2021)
- [Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models](https://mlsanity.com/paper/2409.17146/) (2024)
- [FigureQA: An Annotated Figure Dataset for Visual Reasoning](https://mlsanity.com/paper/1710.07300/) (2017)
- [What's "up" with vision-language models? Investigating their struggle with spatial reasoning](https://mlsanity.com/paper/2310.19785/) (2023)

## Built on by (most cited first, 7 of 7)

- [MMRPT: MultiModal Reinforcement Pre-Training via Masked Vision-Dependent Reasoning](https://mlsanity.com/paper/2512.07203/) (2025)
- [Temporal Self-Distillation: Learning Visual State Tracking in Videos Without Supervision](https://mlsanity.com/paper/2609.04203/) (2026)
- [Logit Lens Supervision for Patch-Level Explanations in Vision-Language Models](https://mlsanity.com/paper/2602.01530/) (2026)
- [Encoded but Not Routed: Explaining the Table-Chart Gap in Scientific Claim Verification](https://mlsanity.com/paper/2606.01679/) (2026)
- [SpatialAfford: Teaching Compact VLMs Where to Look and Where to Ground for Affordance](https://mlsanity.com/paper/2608.00502/) (2026)
- [Where To Look? : Causal Tracing of Vision Encoders in VLM](https://mlsanity.com/paper/2608.10758/) (2026)
- [Look Where You Say You're Looking: Self-Grounded Attention for Visual Reasoning](https://mlsanity.com/paper/2610.05023/) (2026)

## Abstract (arXiv)

Vision Language Models (VLMs) mix visual tokens and text tokens. A puzzling issue is the fact that visual tokens most related to the query receive little to no attention in the final layers of the LLM module of VLMs from the answer tokens, where all tokens are treated equally, in particular, visual and language tokens in the LLM attention layers. This fact may result in wrong answers to visual questions, as our experimental results confirm. It appears that the standard next-token prediction (NTP) loss provides an insufficient signal for directing attention to visual tokens. We hypothesize that a more direct supervision of the attention of visual tokens to corresponding language tokens in the LLM module of VLMs will lead to improved performance on visual tasks. To demonstrate that this is indeed the case, we propose a novel loss function that directly supervises the attention of visual tokens. It directly grounds the answer language tokens in images by directing their attention to the relevant visual tokens. This is achieved by aligning the attention distribution of visual tokens to ground truth attention maps with KL divergence. The ground truth attention maps are obtained from task geometry in synthetic cases or from standard grounding annotations (e.g., bounding boxes or point annotations) in real images, and are used inside the LLM for attention supervision without requiring new labels. The obtained KL attention loss (KLAL) when combined with NTP encourages VLMs to attend to relevant visual tokens while generating answer tokens. This results in notable improvements across geometric tasks, pointing, and referring expression comprehension on both synthetic and real-world data, as demonstrated by our experiments. We also introduce a new dataset to evaluate the line tracing abilities of VLMs. Surprisingly, even commercial VLMs do not perform well on this task.