Exposure Bias Can Alleviate Itself via Directional and Frequency Rectification in Flow Matching
Evolving story · 1 updatesDEFAR: Mitigating Exposure Bias in Flow MatchingTimeline →Researchers propose DEFAR, a method to mitigate exposure bias in Flow Matching by dynamically simulating inference during training, using directional and frequency rectification without external heuristics.
Flow Matching (FM) has achieved remarkable generative performance, yet it suffers from exposure bias due to discrepancies between training and inference. Existing mitigation strategies typically rely on static constraints or external heuristics. In this work, we propose that exposure bias itself inherently contains dynamic signals that can guide its own rectification. To leverage this, we introduce DEFAR (DirEctional-Frequency Adaptive Rectification). This framework simulates the single-step inference process during training to identify exposure bias. It utilizes directional and frequency-adap
Source: Exposure Bias Can Alleviate Itself via Directional and Frequency Rectification in Flow Matching. Read the full piece at the source.
Summary and analysis generated by AI (mistral). Always verify against the original sources.