CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty
We introduce Continuous Adversarial Training for Consistent Honesty (CATCH), inspired by a method originally developed to robustly refuse harmful queries.
ProofPaper ↗
Key points
- Large language models (LLMs) are becoming prevalent in scenarios where honesty is critical.
- CATCH trains the model towards an honest answer and away from a dishonest one while maintaining the model's capabilities.
- We show that trained with only hundreds of targets built from self-reporting scenarios, CATCH learns an honest behaviour that robustly generalizes and transfers across distinct benchmarks.
- We show that CATCH outperforms prompting and supervised fine-tuning in improving honesty, and it is on par with state-of-the-art honesty activation steering.
Sources (1)
- [1]CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM HonestyarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 05:43 PM
We introduce Continuous Adversarial Training for Consistent Honesty (CATCH), inspired by a method originally developed to robustly refuse harmful queries.
Large language models (LLMs) are becoming prevalent in scenarios where honesty is critical.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 15, 2026Decomposing Brain-LLM Alignment with Interpretable Semantic Representations
- Oct 15, 2026FlashAttention with Statistics: Low-Overhead Attention Features for Compact Hallucination Detectors
- Oct 15, 2026Incentivizing Tool-Use Reinforcement Learning for Post-Arrival Emergency Care Prediction
- Oct 15, 2026Fractional Power-of-Two Quantization for Efficient and Effective Multiplier-Free LLM Inference
- Oct 15, 2026Librarian Agents: Training LLMs to Curate Long-Context Workspaces that Anticipate Future Queries
- Oct 14, 2026Optimal Hydra: Model Heterogeneous Federated Learning for LLMs using Data-Informed Hydra-Nets