paper

Posterior Refinement: Fast Language Generation via Any-Order Flow Maps

paperactiveprovisionalposterior-refinement-fast-language-generation-via-any-order-flow-maps-9210a5e0·1 events·first seen 17h ago

Aliases: Posterior Refinement: Fast Language Generation via Any-Order Flow Maps

Co-occurring entities

Posterior Refinement Flow Map Language Models FMLM+Masked Diffusion Models

More like this (12)

Conditional Generative Flow Networks Flow Map Language Models Language Generation in the Limit FlowEdit: Associative Memory for Lifelong Pronunciation Adaptation in Flow-Matching TTS generative language modeling Transformer Language Models Langflow Reasoning Language Models Speaking the Language of Science: Toward a General-Purpose Generative Foundation Model for the Natural Sciences Language Model Finetuning multi-turn language models Posterior Refinement

Recent events (1)

5arXiv · cs.CL·17h ago·source ↗

FMLM+ introduces Posterior Refinement for fast non-autoregressive language generation

Researchers introduce FMLM+, a framework combining Flow Map Language Models with masking-style noise schedules to enable joint sequence generation with per-token global consistency scoring. The key contribution is Posterior Refinement, an inference-time self-correction strategy that matches discrete baseline performance with 32x fewer neural function evaluations (NFEs). The approach improves the speed-quality tradeoff over both Masked Diffusion Models and standard FLMMs across multiple benchmarks, addressing longstanding factorization error problems in non-autoregressive generation.

Frontier Model Releases Inference Economics Posterior Refinement Flow Map Language Models FMLM++2 more