Trajectory matching trains discrete flows on structured action sequences
Authors report MOSES validity rising from 92.8% to 99.2% versus factorized discrete flow matching, with transfer without retraining.
Stories per day · 30 days
Relatedattention1discrete-flow-matching1flexattention1generative-models1gpu-memory1molecule-generation1pytorch1
4 stories, newest first · one card per story
Stories tagged #graph-generation
Edge-aware attention cuts graph-generation GPU memory about 65%
A FlexAttention implementation kept generation quality near the baseline while improving training and sampling throughput.
MAGIC: Mixed-Granularity Agent Graphs via Incremental Construction with Dense-Reward Reinforcement Learning
MAGIC uses dense-reward reinforcement learning to build mixed-granularity multi-agent graphs that beat baselines on eight benchmarks.
Embedded Graph Flows for Categorical Graph Generation
Researchers propose Embedded Graph Flows, a generative model with learned categorical embeddings that beats DiGress and GruM on molecular graph benchmarks.