Learning Cross-Model Activation Alignments with Explicit Many-to-Many Layer Maps
LLMs are released at a rapid pace, raising a natural question: how do two independently trained models relate, both in which layers correspond and in how features transform between them?
ProofPaper ↗
Key points
- We study this by learning an activation alignment, a map from a source model's layerwise activations to a target's.
- Our method, MATCHA, factors this map into a layer map, whose output is an explicit target-by-source matrix that can be extracted and inspected, and a layer-shared feature map between hidden spaces.
- Most of prior work fixes the layer correspondence in advance, pairing layers at roughly the same relative depth; in contrast, we learn both factors jointly from prompts.
- Across 42 pairs of seven models spanning three different families, MATCHA reconstructs the target's activations more faithfully and improves retrieval-based metrics substantially, w.r.t. previous approaches.
Sources (1)
- [1]Learning Cross-Model Activation Alignments with Explicit Many-to-Many Layer MapsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 08:08 PM
LLMs are released at a rapid pace, raising a natural question: how do two independently trained models relate, both in which layers correspond and in how features transform between them?
We study this by learning an activation alignment, a map from a source model's layerwise activations to a target's.
Extractive summary: sentences quoted from the sources.