How to post-train on a surrogate: Envelope sampling mitigates reward hacking
cs.LG, cs.AI, stat.ME
Submitted: 2026-10-08
Updated: 2026-10-08
Code: https://github.com/sanjitdp/envelope-sampling
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- Proximal Policy Optimization Algorithms
- Categorizing Variants of Goodhart's Law
- Reward Hacking in Rubric-Based Reinforcement Learning
- Reproducing, Analyzing, and Detecting Reward Hacking in Rubric-Based Reinforcement Learning
- Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges
- Fine-Tuning Language Models from Human Preferences
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Constitutional AI: Harmlessness from AI Feedback
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- Reward Model Overoptimisation in Iterated RLHF
- ActiveUltraFeedback: Efficient Preference Data Generation using Active Learning
- Uncertainty-Penalized Reinforcement Learning from Human Feedback with Diverse Reward LoRA Ensembles
- Alternating Reinforcement Learning for Rubric-Based Reward Modeling in Non-Verifiable LLM Post-Training
- Rethinking Rubric Generation for Improving LLM Judge and Reward Modeling for Open-ended Tasks
- Qwen3 Technical Report
- Qwen2.5 Technical Report
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