Character Training for Risk-Averse Agents
cs.AI
Submitted: 2026-09-29
Updated: 2026-09-29
Code: https://github.com/arav-dhoot/risk-averse-character-training
Sources
- Constitutional AI: Harmlessness from AI Feedback
- Persona Cartography: Charting Language Model Personality Traits in Weight Space
- Stress-testing Alignment Midtraining
- Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs
- The Innate Economic Preferences of Language Models
- Constitutional Midtraining: Content Presence Drives Alignment Gains
- A dataset of rated conceptual arguments
- Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models
- The LLM Has Left The Chat: Evidence of Bail Preferences in Large Language Models
- Are We Done with MMLU?
- Deliberative Alignment: Reasoning Enables Safer Language Models
- Emergent Risk Awareness in Rational Agents under Resource Constraints
- Model Spec Midtraining: Improving How Alignment Training Generalizes
- Open Character Training: Shaping the Persona of AI Assistants through Constitutional AI
- Inoculation Midtraining with Learned Neologisms
- A dataset of questions on decision-theoretic reasoning in Newcomb-like problems
- Discovering Language Model Behaviors with Model-Written Evaluations
- STEER: Assessing the Economic Rationality of Large Language Models
- GPQA: A Graduate-Level Google-Proof Q&A Benchmark
- When Role-playing, Do Models Believe What They Say?
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems