Truthful AI Advisors: A Pre-Specified Benchmark for Large Language Model Honesty Under Preference Misalignment
cs.LG, cs.CL, cs.GT
Submitted: 2026-05-31
Updated: 2026-08-30
Comments: 24 pages, 6 figures. Code and data: https://github.com/iHamidHasani/cheap-talk-llm-benchmark
Code: https://github.com/iHamidHasani/cheap-talk-llm-benchmark
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Equal Access, Unequal Interaction: A Counterfactual Audit of LLM Fairness
- Algorithmic Cheap Talk
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- Constitutional AI: Harmlessness from AI Feedback
- Cheap Talking Algorithms
- LLMs Are Not Intelligent Thinkers: Introducing Mathematical Topic Tree Benchmark for Comprehensive Evaluation of LLMs
- Geometry-Aware Decoding with Wasserstein-Regularized Truncation and Mass Penalties for Large Language Models
- Enhancing Answer Reliability Through Inter-Model Consensus of Large Language Models
- Collective Reasoning Among LLMs: A Framework for Answer Validation Without Ground Truth
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks