AI/ML●●Solid
AST-guard A gradient-immune structural guard against RL reward hacking
Gradient-immune AST analysis that RL models can't optimize against through backpropagation.
Big BrainNiche Gem
thinking-nick
3019d ago
Back to browse
GitHub Repository
Plug-and-play reward monitoring for RL training loops. Catch reward hacking, component imbalance, and starvation before they tank your run. Drop in one .step() call — get balance reports, auto weight correction, alignment scores, and WandB/TensorBoard/SB3 integrations out of the box. → rewardguard.dev
6 starsPython
RewardGuard – detect reward hacking in RL training loops
by Giovan321·Apr 26, 2026·1 point·1 comment
AI Analysis
●●SolidNiche GemBig Brain
Catches reward hacking before it tanks your RL training run.
Strengths
- •Targets reward hacking specifically, a genuinely hard RL debugging problem
- •Integrates with existing RL stack: WandB, TensorBoard, Stable Baselines3
- •Clear output format with actionable weight adjustment recommendations
Weaknesses
- •Premium auto-adjustment features are private, can't evaluate full value
- •Very early stage: 2 stars, zero issues or pull requests on GitHub
Category
Target Audience
Reinforcement learning engineers and ML researchers
Similar To
Weights & Biases · TensorBoard · MLflow
Similar Projects
Security●●Solid
Argus – Self-hosted Ethereum security monitor
Post-deployment monitoring fills gap that Slither and Mythril leave open for live chains.
WizardryBold Bet
cd4761
114mo ago
AI/ML●●●Banger
RewardHackWatch – Reward hacking detector for LLM agents
Catches LLM reward hacking at runtime when models game evals.
Big BrainWizardryShip It
aerosta
114mo ago
Education●Mid
rlvrbook
Educational content in a space where Nathan Lambert's RLHF book already exists.
Niche Gem
kyars
113mo ago
Developer Tools●●Solid
Squawk – Detect and stop behavioral anti-patterns in AI coding agents
Stateful pattern detection across multiple actions where single-event hooks fail.
Solve My ProblemShip It
jack-lin
214mo ago
AI/ML●●Solid
Agent Wellbeing Kit – boundary protection for humans running AI agents
Error registry catches stuck agent loops before they waste hours of compute.
Big BrainNiche Gem
joozio
103mo ago