# Lost at the End: Primacy Bias in Multimodal Retrieval-Augmented Question Answering

arXiv: https://arxiv.org/abs/2606.16494 · HTML: https://mlsanity.com/paper/2606.16494/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal RAG, one of 15 directions in Retrieval-augmented generation.
- Builds on 19 of its 32 references that are in the map.
- First posted on arXiv 15 Jun 2026.

## Placed in

- Retrieval-augmented generation / [Multimodal RAG](https://mlsanity.com/rag/multimodal-rag/)

## Builds on (most cited first, 19 of 19)

- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT](https://mlsanity.com/paper/2004.12832/) (2020)
- [Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity](https://mlsanity.com/paper/2104.08786/) (2021)
- [Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation](https://mlsanity.com/paper/2108.12409/) (2021)
- [Can Pre-trained Vision and Language Models Answer Visual Information-Seeking Questions?](https://mlsanity.com/paper/2302.11713/) (2023)
- [Encyclopedic VQA: Visual questions about detailed properties of fine-grained categories](https://mlsanity.com/paper/2306.09224/) (2023)
- [Found in the Middle: Calibrating Positional Attention Bias Improves Long Context Utilization](https://mlsanity.com/paper/2406.16008/) (2024)
- [Wiki-LLaVA: Hierarchical Retrieval-Augmented Generation for Multimodal LLMs](https://mlsanity.com/paper/2404.15406/) (2024)
- [Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankers](https://mlsanity.com/paper/2410.02642/) (2024)
- [PreFLMR: Scaling Up Fine-Grained Late-Interaction Multi-modal Retrievers](https://mlsanity.com/paper/2402.08327/) (2024)
- [Order Matters: Exploring Order Sensitivity in Multimodal Large Language Models](https://mlsanity.com/paper/2410.16983/) (2024)
- [According to Me: Long-Term Personalized Referential Memory QA](https://mlsanity.com/paper/2603.01990/) (2026)
- [Do RAG Systems Really Suffer From Positional Bias?](https://mlsanity.com/paper/2505.15561/) (2025)
- [Who is in the Spotlight: The Hidden Bias Undermining Multimodal Retrieval-Augmented Generation](https://mlsanity.com/paper/2506.11063/) (2025)
- [Positional Bias in Multimodal Embedding Models: Do They Favor the Beginning, the Middle, or the End?](https://mlsanity.com/paper/2511.11216/) (2025)

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

- [Multi-Granularity Context-Enhanced RAG over Multimodal Knowledge Graphs](https://mlsanity.com/paper/2608.25986/) (2026)

## Abstract (arXiv)

Knowledge-based visual question answering (KB-VQA) lets vision-language systems answer questions that exceed their parametric knowledge by conditioning a reader on passages retrieved from a Wikipedia-derived knowledge base. In pure-text long-context LLMs, retrieved-context use follows the U-shaped lost-in-the-middle effect of Liu et al. (2024): information at the start and end of context is used, the middle is lost. Whether this transfers to deployed multimodal KB-VQA is open. To close this gap, we design the first controlled probe of reader-side position dependence in multimodal KB-VQA: a gold-position protocol in which only the gold passage's prompt slot varies within question. We run it on three open-source 7B/8B VLM readers and two KB-VQA benchmarks with up to 20 retrieved passages. The shape flips from U to primacy: gold-at-first beats gold-at-last by 16 to 26 points on all six combinations of reader and benchmark, an effect we call Lost at the End; the gap holds at every scale we test, 3B to 32B, attenuating at 32B. Three targeted ablations narrow the cause. A text-only control that removes the image and changes nothing else shows the primacy is already present in text mode and does not depend on the image. Image-position and distractor-shuffle ablations trace the effect to prompt slot 0 of the instruction-tuned reader, where a second answer-bearing passage placed later is largely wasted. On a frozen reader, three retrieval-side fixes (MMR, oracle reranking, rank-based reordering) all fail to improve on the deployment default. Our findings indicate that recall@k is the wrong metric for deployed KB-VQA and that the remaining headroom sits on the reader side; we release our protocol as a controlled instrument for evaluating reader-side interventions.