method
active
method:toy-model-of-crosscoder-diffingToy model of crosscoder diffing
Synthetic two-model activations built from shared and exclusive latent factors with controllable ground truth.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Insights on Crosscoder Model Diffingintroducesmentions
Methods (1)
method
- Crosscoder model diffingrelated_toTrain a single SAE encoding and decoding activations from two models simultaneously to find shared and exclusive features.
Claims (1)
claim
- Central interpretation.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Conclusion.
- Crosscoder model diffing is best used as a high-recall pre-screening tool followed by verification.claim0.816Recommended workflow.
- Extension of model diffing to models with different architectures.
- Approach to identify interpretable differences between a base LLM and its fine-tuned version using SAE latents
- The paper's primary mechanistic analysis method: comparing SAE latent activations before and after fine-tuning to identify misalignment-relevant features
- Validates toy model.
- A priori unknown behavior discovered.
- Learn a shared feature space across two models; basis of the work.