method
active
method:aos-as-model-diffing-techniqueAOs as model diffing technique
Injecting the activation difference between fine-tuned and original model into the AO.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Task of discovering a fine-tuned model has become emergently misaligned, scored on a 0-5 rubric.
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.
- Approach to identify interpretable differences between a base LLM and its fine-tuned version using SAE latents
- Overall positioning.
- The paper's primary mechanistic analysis method: comparing SAE latent activations before and after fine-tuning to identify misalignment-relevant features
- Core motivation.
- Argument for scalability.
- Training procedure for AOs on activation QA data.
- Proposed future direction.
- Scaling result over Qwen-3 8B, Gemma-2 9B, Llama-3.3 70B, Claude Haiku 3.5.