AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Trajectory balance: Improved credit assignment in GFlowNets

Nikolay Malkin, Moksh Jain, Emmanuel Bengio, Chen Sun, Yoshua Bengio

NeurIPS 2022 · Reinforcement learning › Policy gradient and actor-critic · approach: Regularized scalable deep value learning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Train with trajectory balance (TB) instead of flow matching / detailed balanceGFlowNet on AMP generation · AMP generationTop 100 Reward: 0.78 -> 0.85Improved
Train with TB instead of FM/DBGFlowNet on AMP generation · AMP generationTop 100 Diversity: 12.61 -> 18.35Improved
Use TB instead of SACGFlowNet on AMP generation · AMP generationTop 100 Reward: 0.80 -> 0.85Improved
Use TB instead of MCMCGFlowNet on AMP generation · AMP generationTop 100 Diversity: 12.56 -> 18.35Improved
Train with TB instead of FMGFlowNet on molecule generation · Molecule generation held-outPearson correlation log R vs log p: 0.31 -> 0.86Improved
Train with TB instead of DBGFlowNet on molecule generation · Molecule generation held-outPearson correlation log R vs log p: 0.8 -> 0.86Improved
Train with TB instead of FMGFlowNet on molecule generation · Molecule generationDiversity of top 1000 samples (Tanimoto similarity)Improved
Train with TB instead of FM (no parent enumeration)GFlowNet on molecule generation · Molecule generationruntime: up to 5x runtime speedupImproved
Train with DB or TB instead of FMGFlowNet on hypergrid · Hypergrid L1 errorconvergence speedImproved
Learn backward policy PB instead of fixing it uniform (ablation)DB and TB GFlowNets · Hypergrid 64 x 64convergence speedImproved
Train with TB instead of FMGFlowNet on bit sequences · Bit sequences n=120, k in {1,2,4,6,8,10}Spearman correlation with rewardImproved
Increase action space size (bits k) with FM trainingGFlowNet FM · Bit sequencesSpearman correlationMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy