Conflict-Aware Additive Guidance (g^car) is a lightweight learnable method that dynamically resolves gradient conflicts to prevent off-manifold drift in compositional guided sampling for flow models.
If ˆx1 lies inside the region where r is large, the approximate guidance is more accurate, as the optimization is conducted locally and the gradient reflects the landscape well
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Conflict-Aware Additive Guidance for Flow Models under Compositional Rewards
Conflict-Aware Additive Guidance (g^car) is a lightweight learnable method that dynamically resolves gradient conflicts to prevent off-manifold drift in compositional guided sampling for flow models.