# Shaping Initial State Prevents Modality Competition in Multi-modal Fusion: A Two-stage Scheduling Framework via Fast Partial Information Decomposition

arXiv: https://arxiv.org/abs/2509.20840 · HTML: https://mlsanity.com/paper/2509.20840/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 45 references that are in the map.
- First posted on arXiv 25 Sep 2025.

## Builds on (most cited first, 23 of 23)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Gradient Surgery for Multi-Task Learning](https://mlsanity.com/paper/2001.06782/) (2020)
- [Multimodal Machine Learning: A Survey and Taxonomy](https://mlsanity.com/paper/1705.09406/) (2017)
- [GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks](https://mlsanity.com/paper/1711.02257/) (2017)
- [Multi-Task Learning as Multi-Objective Optimization](https://mlsanity.com/paper/1810.04650/) (2018)
- [What Makes Training Multi-Modal Classification Networks Hard?](https://mlsanity.com/paper/1905.12681/) (2019)
- [Balanced Multimodal Learning via On-the-fly Gradient Modulation](https://mlsanity.com/paper/2203.15332/) (2022)
- [Learning Not to Learn: Training Deep Neural Networks with Biased Data](https://mlsanity.com/paper/1812.10352/) (2018)
- [Modality Competition: What Makes Joint Training of Multi-modal Network Fail in Deep Learning? (Provably)](https://mlsanity.com/paper/2203.12221/) (2022)
- [PMR: Prototypical Modal Rebalance for Multimodal Learning](https://mlsanity.com/paper/2211.07089/) (2022)
- [Quantifying & Modeling Multimodal Interactions: An Information Decomposition Framework](https://mlsanity.com/paper/2302.12247/) (2023)
- [Multi-modal Sensor Fusion for Auto Driving Perception: A Survey](https://mlsanity.com/paper/2202.02703/) (2022)
- [Boosting Multi-modal Model Performance with Adaptive Gradient Modulation](https://mlsanity.com/paper/2308.07686/) (2023)
- [MMPareto: Boosting Multimodal Learning with Innocent Unimodal Assistance](https://mlsanity.com/paper/2405.17730/) (2024)
- [ReconBoost: Boosting Can Achieve Modality Reconcilement](https://mlsanity.com/paper/2405.09321/) (2024)
- [On Uni-Modal Feature Learning in Supervised Multi-Modal Learning](https://mlsanity.com/paper/2305.01233/) (2023)
- [Enhancing multimodal cooperation via sample-level modality valuation](https://mlsanity.com/paper/2309.06255/) (2023)
- [Diagnosing and Re-learning for Balanced Multimodal Learning](https://mlsanity.com/paper/2407.09705/) (2024)
- [MetaBalance: Improving Multi-Task Recommendations via Adapting Gradient Magnitudes of Auxiliary Tasks](https://mlsanity.com/paper/2203.06801/) (2022)
- [On-the-fly Modulation for Balanced Multimodal Learning](https://mlsanity.com/paper/2410.11582/) (2024)
- [MMCosine: Multi-Modal Cosine Loss Towards Balanced Audio-Visual Fine-Grained Learning](https://mlsanity.com/paper/2303.05338/) (2023)
- [Learning to Rebalance Multi-Modal Optimization by Adaptively Masking Subnetworks](https://mlsanity.com/paper/2404.08347/) (2024)

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

- [DISC: Decoupling Instruction from State-Conditioned Control via Policy Generation](https://mlsanity.com/paper/2605.20856/) (2026)

## Abstract (arXiv)

Multi-modal fusion often suffers from modality competition during joint training, where one modality dominates the learning process, leaving others under-optimized. Overlooking the critical impact of the model's initial state, most existing methods address this issue during the joint learning stage. In this study, we introduce a two-stage training framework to shape the initial states through unimodal training before the joint training. First, we propose the concept of Effective Competitive Strength (ECS) to quantify a modality's competitive strength. Our theoretical analysis further reveals that properly shaping the initial ECS by unimodal training achieves a provably tighter error bound. However, ECS is computationally intractable in deep neural networks. To bridge this gap, we develop a framework comprising two core components: a fine-grained computable diagnostic metric and an asynchronous training controller. For the metric, we first prove that mutual information(MI) is a principled proxy for ECS. Considering MI is induced by per-modality marginals and thus treats each modality in isolation, we further propose FastPID, a computationally efficient and differentiable solver for partial information decomposition, which decomposes the joint distribution's information into fine-grained measurements: modality-specific uniqueness, redundancy, and synergy. Guided by these measurements, our asynchronous controller dynamically balances modalities by monitoring uniqueness and locates the ideal initial state to start joint training by tracking peak synergy. Experiments on diverse benchmarks demonstrate that our method achieves state-of-the-art performance. Our work establishes that shaping the pre-fusion models' initial state is a powerful strategy that eases competition before it starts, reliably unlocking synergistic multi-modal fusion.