# Teaching Vision-Language Models to Decline: Support Completion Against Mirage

arXiv: https://arxiv.org/abs/2610.16465 · HTML: https://mlsanity.com/paper/2610.16465/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 24 of its 30 references that are in the map.
- First posted on arXiv 12 Oct 2026.

## Placed in

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

## Builds on (most cited first, 24 of 24)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks](https://mlsanity.com/paper/1610.02136/) (2016)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency](https://mlsanity.com/paper/2508.18265/) (2025)
- [A General Language Assistant as a Laboratory for Alignment](https://mlsanity.com/paper/2112.00861/) (2021)
- [VizWiz Grand Challenge: Answering Visual Questions from Blind People](https://mlsanity.com/paper/1802.08218/) (2018)
- [Selective Classification for Deep Neural Networks](https://mlsanity.com/paper/1705.08500/) (2017)
- [Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding](https://mlsanity.com/paper/2311.16922/) (2023)
- [Gradient Starvation: A Learning Proclivity in Neural Networks](https://mlsanity.com/paper/2011.09468/) (2020)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [Measuring Progress on Scalable Oversight for Large Language Models](https://mlsanity.com/paper/2211.03540/) (2022)
- [Truthful AI: Developing and governing AI that does not lie](https://mlsanity.com/paper/2110.06674/) (2021)
- [R-Tuning: Instructing Large Language Models to Say `I Don't Know'](https://mlsanity.com/paper/2311.09677/) (2023)
- [Reliable Visual Question Answering: Abstain Rather Than Answer Incorrectly](https://mlsanity.com/paper/2204.13631/) (2022)
- [MIRAGE: The Illusion of Visual Understanding](https://mlsanity.com/paper/2603.21687/) (2026)
- [Mirage Probes: How Vision Models Fake Visual Understanding](https://mlsanity.com/paper/2606.13870/) (2026)
- [Detect Before You Leap: Mirage Detection in Vision-Language Models](https://mlsanity.com/paper/2606.00435/) (2026)

## Abstract (arXiv)

Vision-language models answer questions about images they were never shown, a failure called mirage. Such an answer can be correct without being justified by the evidence, and the difference matters wherever a model's output is checked by someone who did not see its input. We locate the failure in the training objective: cross-entropy separates a model that reads the image from one that ignores it by only the conditional mutual information between answer and image, and no training example rewards declining. We formalise support completion, TIDE, which also shows each training question with a blank image, a same-domain wrong image and an out-of-domain image, with targets that decline, and prove that under TIDE every image-blind predictor is suboptimal by at least 1.04 nats per example, that the optimum declines exactly on the counterfactual kinds while keeping its behaviour on valid images, and that for any finite model the error of its answer-or-decline decision, declining on valid input plus answering on invalid input, is bounded below by how well its representation separates valid from invalid evidence. On Qwen2.5-VL-7B under a frozen held-out protocol with two judges, the three-condition predecessor of TIDE reduced the invalid-evidence mirage rate from 0.759 to 0.097 and the blank-image rate from 0.700 to 0.003 at a false-abstention cost of 0.126; label-free grounding raised the rate to 0.859, a likelihood-advantage contrast moved it only to 0.718 because its empty-image term carries no gradient, and grounding applied after support completion re-opened the mirage. The four-condition protocol on three base models, with seeds and component ablations, is in progress and not reported here.