What Linear Probes Miss: Multi-View Probing for Weight-Space Learning
About
The explosive growth of open-source model repositories has created a Model Jungle, where checkpoints are frequently shared without adequate documentation or metadata. While weight-space learning offers a pathway to identify and analyze these models directly from their parameters, processing full-scale weights is computationally prohibitive. Probing-based methods have emerged as a lightweight alternative, extracting permutation-equivariant representations via learnable probe vectors. However, existing probing methods are limited by a single-view design: they capture first-order structures but fail to encode the rich, higher-order correlation patterns inherent in row-column interactions. To bridge this gap, we introduce MVProbe, a multi-perspective probing framework that synthesizes first-order signals with interaction-aware (Gram-based) views. Our approach is theoretically grounded; we analyze the scaling laws of different probing orders to derive a principled standardization and fusion strategy that ensures balanced contributions from all branches. On the Model Jungle benchmark, MVProbe consistently outperforms the state-of-the-art ProbeX across diverse architectures, including discriminative backbones (ResNet, SupViT, MAE, DINO) and large-scale generative LoRA adapters (Stable Diffusion LoRA).
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Weight-Space Classification | Model Jungle | Accuracy92.33 | 20 | |
| Model Classification | SD200 zero-shot (holdout) | Accuracy95.53 | 7 | |
| kNN retrieval | SD LoRA (SD200) | Accuracy99.8 | 6 | |
| kNN retrieval | SD LoRA (SD1k) | Accuracy93.99 | 6 | |
| One-Class Classification (OCC) | SD LoRA (SD200) | Accuracy100 | 6 | |
| Weight-Space Classification | MNIST INR | Accuracy97.2 | 5 | |
| One-Class Classification (OCC) | SD LoRA (SD1k) | Accuracy99.99 | 4 | |
| Weight-Space Classification | Stable Diffusion LoRA In-Distribution SD200 (train) | Accuracy99.8 | 2 | |
| Weight-Space Classification | Stable Diffusion LoRA In-Distribution SD1k (train) | Accuracy97.88 | 2 | |
| Weight-Space Classification | Stable Diffusion LoRA Zero-shot SD1k (held-out classes) | Accuracy97.96 | 2 |