Edit Flows: Flow Matching with Edit Operations
By: Marton Havasi , Brian Karrer , Itai Gat and more
Potential Business Impact:
Lets computers write better stories and code.
Autoregressive generative models naturally generate variable-length sequences, while non-autoregressive models struggle, often imposing rigid, token-wise structures. We propose Edit Flows, a non-autoregressive model that overcomes these limitations by defining a discrete flow over sequences through edit operations-insertions, deletions, and substitutions. By modeling these operations within a Continuous-time Markov Chain over the sequence space, Edit Flows enable flexible, position-relative generation that aligns more closely with the structure of sequence data. Our training method leverages an expanded state space with auxiliary variables, making the learning process efficient and tractable. Empirical results show that Edit Flows outperforms both autoregressive and mask models on image captioning and significantly outperforms the mask construction in text and code generation.
Similar Papers
Edit-Based Flow Matching for Temporal Point Processes
Machine Learning (CS)
Makes predicting event times faster and more flexible.
Edit-Based Flow Matching for Temporal Point Processes
Machine Learning (CS)
Makes predicting event times faster and more accurate.
Flowception: Temporally Expansive Flow Matching for Video Generation
CV and Pattern Recognition
Makes videos by adding and fixing pictures.