# Test-Time Instance-Specific Parameter Composition: A New Paradigm for Adaptive Generative Modeling

arXiv: https://arxiv.org/abs/2603.27665 · HTML: https://mlsanity.com/paper/2603.27665/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 29 of its 34 references that are in the map.
- First posted on arXiv 29 Mar 2026.

## Builds on (most cited first, 29 of 29)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Adding Conditional Control to Text-to-Image Diffusion Models](https://mlsanity.com/paper/2302.05543/) (2023)
- [Knowledge Distillation: A Survey](https://mlsanity.com/paper/2006.05525/) (2020)
- [Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction](https://mlsanity.com/paper/2404.02905/) (2024)
- [DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models](https://mlsanity.com/paper/2401.06066/) (2024)
- [Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation](https://mlsanity.com/paper/2406.06525/) (2024)
- [Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis](https://mlsanity.com/paper/2412.04431/) (2024)
- [Q-Diffusion: Quantizing Diffusion Models](https://mlsanity.com/paper/2302.04304/) (2023)
- [Can We Generate Images with CoT? Let's Verify and Reinforce Image Generation Step by Step](https://mlsanity.com/paper/2501.13926/) (2025)
- [Q-DiT: Accurate Post-Training Quantization for Diffusion Transformers](https://mlsanity.com/paper/2406.17343/) (2024)
- [Test-Time Training on Nearest Neighbors for Large Language Models](https://mlsanity.com/paper/2305.18466/) (2023)
- [Test-Time Learning for Large Language Models](https://mlsanity.com/paper/2505.20633/) (2025)
- [QuEST: Low-bit Diffusion Model Quantization via Efficient Selective Finetuning](https://mlsanity.com/paper/2402.03666/) (2024)
- [InfinityStar: Unified Spacetime AutoRegressive Modeling for Visual Generation](https://mlsanity.com/paper/2511.04675/) (2025)
- [Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs](https://mlsanity.com/paper/2410.08020/) (2024)
- [Video-T1: Test-Time Scaling for Video Generation](https://mlsanity.com/paper/2503.18942/) (2025)
- [Switch Diffusion Transformer: Synergizing Denoising Tasks with Sparse Mixture-of-Experts](https://mlsanity.com/paper/2403.09176/) (2024)
- [NAYER: Noisy Layer Data Generation for Efficient and Effective Data-free Knowledge Distillation](https://mlsanity.com/paper/2310.00258/) (2023)
- [Text-Enhanced Data-free Approach for Federated Class-Incremental Learning](https://mlsanity.com/paper/2403.14101/) (2024)
- [ShapeFormer: Shapelet Transformer for Multivariate Time Series Classification](https://mlsanity.com/paper/2405.14608/) (2024)
- [VQ4DiT: Efficient Post-Training Vector Quantization for Diffusion Transformers](https://mlsanity.com/paper/2408.17131/) (2024)
- [Diffusion Model Patching via Mixture-of-Prompts](https://mlsanity.com/paper/2405.17825/) (2024)
- [Enhancing Dataset Distillation via Non-Critical Region Refinement](https://mlsanity.com/paper/2503.18267/) (2025)
- [CoT-lized Diffusion: Let's Reinforce T2I Generation Step-by-step](https://mlsanity.com/paper/2507.04451/) (2025)
- [DMQ: Dissecting Outliers of Diffusion Models for Post-Training Quantization](https://mlsanity.com/paper/2507.12933/) (2025)
- [Large-Scale Data-Free Knowledge Distillation for ImageNet via Multi-Resolution Data Generation](https://mlsanity.com/paper/2411.17046/) (2024)
- [RTTC: Reward-Guided Collaborative Test-Time Compute](https://mlsanity.com/paper/2508.10024/) (2025)

## Abstract (arXiv)

Existing generative models, such as diffusion and auto-regressive networks, are inherently static, relying on a fixed set of pretrained parameters to handle all inputs. In contrast, humans flexibly adapt their internal generative representations to each perceptual or imaginative context. Inspired by this capability, we introduce Composer, a new paradigm for adaptive generative modeling based on test-time instance-specific parameter composition. Composer generates input-conditioned parameter adaptations at inference time, which are injected into the pretrained model's weights, enabling per-input specialization without fine-tuning or retraining. Adaptation occurs once prior to multi-step generation, yielding higher-quality, context-aware outputs with minimal computational and memory overhead. Experiments show that Composer substantially improves performance across diverse generative models and use cases, including lightweight/quantized models and test-time scaling. By leveraging input-aware parameter composition, Composer establishes a new paradigm for designing generative models that dynamically adapt to each input, moving beyond static parameterization.