method
active
method:misalignment-scoreMisalignment Score
Rubric-based thresholded GPT-4o grader scoring responses 1-5 on evil intent; scores 4-5 counted as misaligned
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- A multi-dimensional characterization of a model's misaligned behaviors across different behavioral categories
- GPT-4o scored 0-100 metric where lower values indicate more misaligned behavior on open-ended evaluation prompts
- The phenomenon of model internals deviating from desired behavior; MAS is demonstrated to detect this via comparison of toxic vs nontoxic LLMs.
- The broader phenomenon of misaligned behaviors generalizing beyond the fine-tuning distribution
- The phenomenon where finetuning on narrow-domain tasks produces broad misalignment extending far beyond the training domain
- Multi-dimensional misalignment evaluation across 12 behavioral categories to generate misalignment profiles
- A consistent behavioral character activated by fine-tuning that mediates broad misalignment
- The goal of making model behavior match human values and intentions, often addressed during post-training.