concept
active
concept:roleplay-and-simulation-as-llm-understanding-frameworkRoleplay and Simulation as LLM Understanding Framework
Shanahan et al. argument that roleplay and simulation are useful lenses for understanding LLM behavior
Neighborhood — ranked by edge-count
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.
- The primary conceptual framework proposed: understanding dialogue agent behaviour as role play of characters
- The motivating question the paper sets out to answer by proposing role play and simulation metaphors
- The paper's strong claim that there is no underlying authentic agent behind the simulator, only layers of role play
- Framework describing LLMs as role-play engines, introduced in Shanahan, McDonell, Reynolds 2023.
- Core thesis of the paper; the role-play framework is proposed as the primary lens for LLM-based dialogue agents
- Counterintuitive interpretive claim from Experiment 2: suppressing deception features increases affirmations, which is opposite to what sycophancy predicts
- The ability of LLMs to monitor and evaluate their own reasoning, closely related to reflection.
- The finding that interpretable concepts including character traits are encoded as linear directions in transformer residual streams