concept
active
concept:model-diffing

Model Diffing

Approach to identify interpretable differences between a base LLM and its fine-tuned version using SAE latents

Neighborhood — ranked by edge-count

Thinkers (3)

thinker
  • Unpublished independent model diffing work.
  • Authors of persona modulation jailbreak dataset used in this paper's evaluations
  • Showed base-model SAEs transfer to finetuned models.

Methods (3)

method
  • Sparse dictionary learning variant that reads and writes to multiple layers or models, yielding shared features.
  • Train a single SAE encoding and decoding activations from two models simultaneously to find shared and exclusive features.
  • Core method: sample from logits_after + α(logits_after − logits_before) to surface rare behaviors.

Concepts (2)

concept

Institutes (1)

institute

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.