# The Ebb and Flow of Multimodal Focus: Scheduling Visual Relay Windows for Grounded VLM Reasoning

arXiv: https://arxiv.org/abs/2607.11436 · HTML: https://mlsanity.com/paper/2607.11436/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 17 of its 64 references that are in the map.
- First posted on arXiv 13 Jul 2026.

## 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, 17 of 17)

- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency](https://mlsanity.com/paper/2508.18265/) (2025)
- [Hallucination of Multimodal Large Language Models: A Survey](https://mlsanity.com/paper/2404.18930/) (2024)
- [A Survey on Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2402.00253/) (2024)
- [See What You Are Told: Visual Attention Sink in Large Multimodal Models](https://mlsanity.com/paper/2503.03321/) (2025)
- [FineVision: Open Data Is All You Need](https://mlsanity.com/paper/2510.17269/) (2025)
- [Seeing Clearly by Layer Two: Enhancing Attention Heads to Alleviate Hallucination in LVLMs](https://mlsanity.com/paper/2411.09968/) (2024)
- [BabyVision: Visual Reasoning Beyond Language](https://mlsanity.com/paper/2601.06521/) (2026)
- [Lost in Embeddings: Information Loss in Vision-Language Models](https://mlsanity.com/paper/2509.11986/) (2025)
- [SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decoding](https://mlsanity.com/paper/2506.08391/) (2025)
- [VIB-Probe: Detecting and Mitigating Hallucinations in Vision-Language Models via Variational Information Bottleneck](https://mlsanity.com/paper/2601.05547/) (2026)
- [The Spatial Blindspot of Vision-Language Models](https://mlsanity.com/paper/2601.09954/) (2026)
- [The Geometry of Representational Failures in Vision Language Models](https://mlsanity.com/paper/2602.07025/) (2026)
- [KVSmooth: Mitigating Hallucination in Multi-modal Large Language Models through Key-Value Smoothing](https://mlsanity.com/paper/2602.04268/) (2026)
- [Large Vision-Language Models Get Lost in Attention](https://mlsanity.com/paper/2605.05668/) (2026)

## Abstract (arXiv)

Vision-language models increasingly succeed on multimodal reasoning benchmarks, yet their visual evidence often becomes unstable once it enters the language stack, weakening evidence-grounded reasoning. To understand this fragility, we examine the internal dynamics of VLMs through a mechanistic lens and uncover a stable three-stage redistribution of multimodal attention focus across depth: an early question-conditioned organization, a critical middle visual-dominant relay, and a late return to answer formation. We operationalize the middle phase as the Visual Relay Window (VRW), and show that its geometry varies with task demand, is causally tied to grounded generation, and distinguishes unsupported answers from stronger reasoning trajectories. Guided by this internal rhythm, we propose TRACE, a task-adaptive inference-time control framework with lightweight trained modules. It reshapes relay allocation during prefill and preserves assembled visual support after handoff during decoding. Across four open-weight VLM backbones and seven benchmarks, TRACE delivers large gains on grounding-sensitive settings, improving them by 4.33 points on average and by up to 6.6 points, while also improving reasoning-heavy tasks. These results show that explicitly controlling multimodal focus across depth offers a unified and effective mechanism for strengthening evidence-grounded multimodal reasoning.