Gradient Regularization Prevents Reward Hacking in Reinforcement Learning from Human Feedback and Verifiable Rewards
Johannes Ackermann, Michael Noukhovitch, Takashi Ishida, Masashi Sugiyama
Abstract
Reinforcement Learning from Human Feedback (RLHF) or Verifiable Rewards (RLVR) are two key steps in the post-training of modern Language Models (LMs). A common problem is reward hacking, where the policy may exploit inaccuracies of the reward and learn an unintended behavior. Most previous works address this by limiting the policy update with a Kullback-Leibler (KL) penalty towards a reference model. We propose a dif...
Results & Benchmarks
Benchmark data is not yet available for this paper.
Hardware Requirements
- Expect multi-day setup/compute for meaningful reproduction based on current guidance.
Best Implementation
Maintained implementation evidence is not confirmed for this paper yet.
Use the Implementation Status and Reproduction Path sections below for the current action plan.
Reproduction Path
Follow this baseline workflow to decide if this paper is worth immediate prototyping.
- 1
No maintained paper-verified implementation was found; start with the closest related repositories below.
- 2
Compare repo methods against the paper equations/algorithm before trusting metrics.
- 3
Create a minimal baseline implementation from the paper and use adjacent repos as references.
- 4
Prioritize reproducing the core method first: Reinforcement learning.
Related Implementations
These are not paper-verified. Use them as reference points when no direct implementation is available.
Matches contextual method/domain keyword: reinforcement learning
Additional Implementations
No additional verified repositories beyond the primary recommendation.
Hugging Face Artifacts
No trustworthy direct or curated related Hugging Face artifacts were found yet.
Continue with targeted Hugging Face searches: