# MemCorr-DP: Counterfactual Correspondence Conditioning for a Diffusion Policy Guided by a Reference

arXiv: https://arxiv.org/abs/2609.06615 · HTML: https://mlsanity.com/paper/2609.06615/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models.
- Builds on 24 of its 32 references that are in the map.
- First posted on arXiv 6 Sep 2026.

## Placed in

- Diffusion models / [Diffusion and flow policies for robotics](https://mlsanity.com/diffusion/diffusion-and-flow-policies-for-robotics/)

## Builds on (most cited first, 24 of 24)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Diffusion Policy: Visuomotor Policy Learning via Action Diffusion](https://mlsanity.com/paper/2303.04137/) (2023)
- [DINOv3](https://mlsanity.com/paper/2508.10104/) (2025)
- [Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning](https://mlsanity.com/paper/1910.10897/) (2019)
- [RoMa: Robust Dense Feature Matching](https://mlsanity.com/paper/2305.15404/) (2023)
- [3D Diffuser Actor: Policy Diffusion with 3D Scene Representations](https://mlsanity.com/paper/2402.10885/) (2024)
- [DKM: Dense Kernelized Feature Matching for Geometry Estimation](https://mlsanity.com/paper/2202.00667/) (2022)
- [Flow as the Cross-Domain Manipulation Interface](https://mlsanity.com/paper/2407.15208/) (2024)
- [EquiBot: SIM(3)-Equivariant Diffusion Policy for Generalizable and Data Efficient Learning](https://mlsanity.com/paper/2407.01479/) (2024)
- [RoMa v2: Harder Better Faster Denser Feature Matching](https://mlsanity.com/paper/2511.15706/) (2025)
- [TAX-Pose: Task-Specific Cross-Pose Estimation for Robot Manipulation](https://mlsanity.com/paper/2211.09325/) (2022)
- [AffordDP: Generalizable Diffusion Policy with Transferable Affordance](https://mlsanity.com/paper/2412.03142/) (2024)
- [GenDP: 3D Semantic Fields for Category-Level Generalizable Diffusion Policy](https://mlsanity.com/paper/2410.17488/) (2024)
- [Render and Diffuse: Aligning Image and Action Spaces for Diffusion-based Behaviour Cloning](https://mlsanity.com/paper/2405.18196/) (2024)
- [CordViP: Correspondence-based Visuomotor Policy for Dexterous Manipulation in Real-World](https://mlsanity.com/paper/2502.08449/) (2025)
- [One-shot Imitation Learning via Interaction Warping](https://mlsanity.com/paper/2306.12392/) (2023)
- [Conditioning Matters: Training Diffusion Policies is Faster Than You Think](https://mlsanity.com/paper/2505.11123/) (2025)
- [A Practical Guide for Incorporating Symmetry in Diffusion Policy](https://mlsanity.com/paper/2505.13431/) (2025)
- [Bridging Perception and Action: Spatially-Grounded Mid-Level Representations for Robot Generalization](https://mlsanity.com/paper/2506.06196/) (2025)
- [Knowledge-Driven Imitation Learning: Enabling Generalization Across Diverse Conditions](https://mlsanity.com/paper/2506.21057/) (2025)
- [Robust Learning of Diffusion Models with Extremely Noisy Conditions](https://mlsanity.com/paper/2510.10149/) (2025)
- [MatchingPolicy: Correspondence-Aware Policy Enables Cross-Object In-Context Learning](https://mlsanity.com/paper/2608.16715/) (2026)

## Abstract (arXiv)

Behavior-cloned visuomotor policies can remain accurate near their training distribution yet fail when object position and camera viewpoint change together. A successful reference trajectory contains the geometry needed to transfer the same interaction, but the policy must align that geometry with the current scene and remain sensitive to it during denoising. To address these challenges, we present MemCorr-DP, a diffusion policy that lifts frozen RoMa v2 matches into explicit 3D relations between the current scene and the reference trajectory. A counterfactual paired objective assigns opposite behaviors the same physical state and noisy action while retaining reference-specific denoising targets. Mixed-condition fine-tuning then adapts the policy from ground-truth geometry to measured correspondence errors. Our strongest evaluation places the Door in the outermost position bands beyond the training support and changes the query camera by ±15^circ. Under this combined shift, MemCorr-DP achieves 96.67% closed-loop success, compared with 88.00% for a visual Transformer with the same action architecture. Objective ablations and reference interventions show that behavior responds to the selected reference, while matched controls favor the complete relation set over future motion or centroid geometry alone. These results support explicit 3D reference relations as a robust conditioning interface when spatial and viewpoint changes are compounded in the evaluated task.