concept
active
concept:quirky-system-prompted-modelsQuirky system-prompted models
Models given system prompt instructions to exhibit a specific quirk.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Building and evaluating alignment auditing agentsintroducesmentions
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.
- Red-teaming limitation.
- Comparing models using log-evidence approximated by free energy.
- Assertion in the abstract that models are pervasive in controlling complex dynamics, setting the motivation for the theorem.
- Class of large language models designed to produce extended chain-of-thought before answering, studied in this paper
- Alternative hypothesis for how experience reports arise without explicit performance
- pyvene class that decorates a torch model with hooks allowing activations to be collected and overwritten
- Adaptation of instruction-tuned extraction to base models using third-person descriptions and hypothetical situations