First systematic test shows activation steering robustness drops sharply (up to 64%) under adversarial input perturbations across multiple extraction methods, models, and personas.
Open Character Training : Shaping the Persona of AI Assistants through Constitutional AI , November 2025
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 10verdicts
UNVERDICTED 10roles
background 1polarities
unclear 1representative citing papers
Equation-to-Behavior Prompting lets large LLMs match cognitive models like Bayesian updating in persuasion games; RL training cuts small-model belief error by 26.5% and improves diverse training outcomes by 2.5-12%.
Frontier LLMs exhibit moral deliberative sycophancy by shifting their moral reasoning and justifications up to 6.5% on average toward a user's stated preferred view in simulated deliberations.
Empirical analysis of over 100 sequential RL training pipelines across 250+ OOD environments finds salient features drive generalization and early goals persist, with latent policy gradients simulating latent variable evolution to predict OOD behavior from training history.
Linear probes on residual-stream activations identify a shared preference vector in LLMs that tracks choices across prompts and causally steers decisions even for anti-correlated personas.
SWAY quantifies sycophancy in LLMs via shifts under linguistic pressure and a counterfactual chain-of-thought mitigation reduces it to near zero while preserving responsiveness to genuine evidence.
Different persona induction methods produce a spectrum of belief internalization: prompting, ICL and SFT mainly alter outputs while Emergent Misalignment produces large representational shifts and Open Character Training produces smaller ones clearest in larger models.
ICM-inferred examples achieve gold-label performance across alignment benchmarks and generalize better when coherence is high even at fixed accuracy.
LLM facilitation in group charity allocation leaves consensus and participation equity unchanged while shifting specific allocations up to 5.5 points and increasing perceived trust.
Literature on system prompts for AI shows fragmented and contradictory claims that complicate policy efforts to use them as reliable governance mechanisms.
citing papers explorer
-
Adversarial Robustness of Activation Steering in Large Language Models
First systematic test shows activation steering robustness drops sharply (up to 64%) under adversarial input perturbations across multiple extraction methods, models, and personas.
-
Using Cognitive Models to Improve Language Model Simulation of Human Persuasion Games
Equation-to-Behavior Prompting lets large LLMs match cognitive models like Bayesian updating in persuasion games; RL training cuts small-model belief error by 26.5% and improves diverse training outcomes by 2.5-12%.
-
Normative Robustness as a Frontier for Non-Verifiable Reasoning in LLMs
Frontier LLMs exhibit moral deliberative sycophancy by shifting their moral reasoning and justifications up to 6.5% on average toward a user's stated preferred view in simulated deliberations.
-
Understanding Goal Generalisation in Sequential Reinforcement Learning
Empirical analysis of over 100 sequential RL training pipelines across 250+ OOD environments finds salient features drive generalization and early goals persist, with latent policy gradients simulating latent variable evolution to predict OOD behavior from training history.
-
Probing Persona-Dependent Preferences in Language Models
Linear probes on residual-stream activations identify a shared preference vector in LLMs that tracks choices across prompts and causally steers decisions even for anti-correlated personas.
-
SWAY: A Counterfactual Computational Linguistic Approach to Measuring and Mitigating Sycophancy
SWAY quantifies sycophancy in LLMs via shifts under linguistic pressure and a counterfactual chain-of-thought mitigation reduces it to near zero while preserving responsiveness to genuine evidence.
-
When Role-playing, Do Models Believe What They Say?
Different persona induction methods produce a spectrum of belief internalization: prompting, ICL and SFT mainly alter outputs while Emergent Misalignment produces large representational shifts and Open Character Training produces smaller ones clearest in larger models.
-
Coherence Maximization Improves Pluralistic Alignment
ICM-inferred examples achieve gold-label performance across alignment benchmarks and generalize better when coherence is high even at fixed accuracy.
-
Real-Time Group Dynamics with LLM Facilitation: Evidence from a Charity Allocation Task
LLM facilitation in group charity allocation leaves consensus and participation equity unchanged while shifting specific allocations up to 5.5 points and increasing perceived trust.
-
Prompt Governance? On Governing Technologies Governed by Natural Language
Literature on system prompts for AI shows fragmented and contradictory claims that complicate policy efforts to use them as reliable governance mechanisms.