# Learning to Orchestrate Vision Foundation Models for Multi-Task Dense Prediction

arXiv: https://arxiv.org/abs/2606.15765 · HTML: https://mlsanity.com/paper/2606.15765/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 29 references that are in the map.
- First posted on arXiv 14 Jun 2026.

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

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [Counterfactual Multi-Agent Policy Gradients](https://mlsanity.com/paper/1705.08926/) (2017)
- [PAD-Net: Multi-Tasks Guided Prediction-and-Distillation Network for Simultaneous Depth Estimation and Scene Parsing](https://mlsanity.com/paper/1805.04409/) (2018)
- [AM-RADIO: Agglomerative Vision Foundation Model -- Reduce All Domains Into One](https://mlsanity.com/paper/2312.06709/) (2023)
- [CroCo: Self-Supervised Pre-training for 3D Vision Tasks by Cross-View Completion](https://mlsanity.com/paper/2210.10716/) (2022)
- [On the Representation Collapse of Sparse Mixture of Experts](https://mlsanity.com/paper/2204.09179/) (2022)
- [InvPT: Inverted Pyramid Multi-task Transformer for Dense Scene Understanding](https://mlsanity.com/paper/2203.07997/) (2022)
- [MoVA: Adapting Mixture of Vision Experts to Multimodal Context](https://mlsanity.com/paper/2404.13046/) (2024)
- [Multi-Task Dense Prediction via Mixture of Low-Rank Experts](https://mlsanity.com/paper/2403.17749/) (2024)
- [TaskExpert: Dynamically Assembling Multi-Task Representations with Memorial Mixture-of-Experts](https://mlsanity.com/paper/2307.15324/) (2023)
- [Routers in Vision Mixture of Experts: An Empirical Study](https://mlsanity.com/paper/2401.15969/) (2024)
- [InvPT++: Inverted Pyramid Multi-Task Transformer for Visual Scene Understanding](https://mlsanity.com/paper/2306.04842/) (2023)
- [TSP-Transformer: Task-Specific Prompts Boosted Transformer for Holistic Scene Understanding](https://mlsanity.com/paper/2311.03427/) (2023)
- [When Are Experts Misrouted? Counterfactual Routing Analysis in Mixture-of-Experts Language Models](https://mlsanity.com/paper/2605.07260/) (2026)
- [Multi-task Learning with 3D-Aware Regularization](https://mlsanity.com/paper/2310.00986/) (2023)

## Abstract (arXiv)

Vision foundation models (VFMs) exhibit complementary strengths shaped by their pretraining objectives. Yet prevailing methods for multi-task dense prediction still train an entire backbone, either by fine-tuning it under multi-task supervision or by distilling multiple VFMs in an additional stage. We ask whether downstream learning can instead compose the frozen representations already available in foundation models. Dense tasks require composite representations that no individual expert provides alone. Realizing them is difficult: simple fusion yields only marginal gains over the best single expert, while learned routing tends to collapse toward candidates that are strong at initialization, starving newly initialized composers of training signal. We present COVE, which constructs pairwise composite candidates through Synergy Composers and routes among raw and composite candidates with a Task-Conditioned Router. To prevent this collapse, COVE combines Gaussian logit perturbation for exploration with counterfactual supervision that selectively increases under-credited routing allocations. On NYUD-v2 and PASCAL-Context, COVE matches or surpasses ViT-L-based methods on most tasks using a smaller frozen encoder pool and roughly half the computation of recent VFM-based competitors, while exceeding the best single frozen expert on every task.