# Rethinking Classifier-Free Guidance for TI2I Flow Matching

Preprint · Andreas Floros, Elena Koritska, Teofilo de Campos, Christoph Vogel, Chu-Tak Li, Anuj Sharma, Alexandros Neophytou

- Archive id: arxiv:2610.21550
- Published: 2026-10-15
- arXiv: https://arxiv.org/abs/2610.21550
- Source URL: https://arxiv.org/abs/2610.21550
- PDF: https://arxiv.org/pdf/2610.21550
- Provenance: arxiv

## Abstract

We study Text-and-Image-to-Image (TI2I) flow matching methods for controllable image editing. In a broader generative modeling context, Classifier-Free Guidance (CFG) is established as a key component responsible for high quality outputs. However, we argue that its role in TI2I modeling is poorly understood. Specifically, we make the surprising observation that the CFG mechanism is implicitly harmful for editing in the noisy phase of flow modeling, hindering instruction following and controllability. We propose a simple correction in the early flow steps, enabling a richer alignment-preservation tradeoff at no extra cost. The effectiveness of our approach is demonstrated on both pixel-space and latent TI2I flows.

_Abstract reproduced under CC0-1.0 (arXiv descriptive metadata)._

---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.