# Causal Negative Sampling via Diffusion Model for Out-of-Distribution Recommendation

arXiv: https://arxiv.org/abs/2508.07243 · HTML: https://mlsanity.com/paper/2508.07243/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Diffusion recommenders, one of 14 directions in Diffusion models.
- Builds on 23 of its 41 references that are in the map.
- First posted on arXiv 10 Aug 2025.

## Placed in

- Diffusion models / [Diffusion recommenders](https://mlsanity.com/diffusion/diffusion-recommenders/)

## Builds on (most cited first, 23 of 23)

- [Neural Collaborative Filtering](https://mlsanity.com/paper/1708.05031/) (2017)
- [Diffusion Models in Vision: A Survey](https://mlsanity.com/paper/2209.04747/) (2022)
- [A Survey on Video Diffusion Models](https://mlsanity.com/paper/2310.10647/) (2023)
- [DiffuRec: A Diffusion Model for Sequential Recommendation](https://mlsanity.com/paper/2304.00686/) (2023)
- [Multi-Task Feature Learning for Knowledge Graph Enhanced Recommendation](https://mlsanity.com/paper/1901.08907/) (2019)
- [Understanding Negative Sampling in Graph Representation Learning](https://mlsanity.com/paper/2005.09863/) (2020)
- [DiffKG: Knowledge Graph Diffusion Model for Recommendation](https://mlsanity.com/paper/2312.16890/) (2023)
- [Plug-in Diffusion Model for Sequential Recommendation](https://mlsanity.com/paper/2401.02913/) (2024)
- [Diffusion Models for Generative Outfit Recommendation](https://mlsanity.com/paper/2402.17279/) (2024)
- [Empowering Collaborative Filtering with Principled Adversarial Contrastive Loss](https://mlsanity.com/paper/2310.18700/) (2023)
- [Negative Sampling for Contrastive Representation Learning: A Review](https://mlsanity.com/paper/2206.00212/) (2022)
- [Fairly Adaptive Negative Sampling for Recommendations](https://mlsanity.com/paper/2302.08266/) (2023)
- [RecDiff: Diffusion Model for Social Recommendation](https://mlsanity.com/paper/2406.01629/) (2024)
- [A Survey on Diffusion Models for Recommender Systems](https://mlsanity.com/paper/2409.05033/) (2024)
- [On the Theories Behind Hard Negative Sampling for Recommendation](https://mlsanity.com/paper/2302.03472/) (2023)
- [Negative Sampling in Knowledge Graph Representation Learning: A Review](https://mlsanity.com/paper/2402.19195/) (2024)
- [Diffusion-based Negative Sampling on Graphs for Link Prediction](https://mlsanity.com/paper/2403.17259/) (2024)
- [Causal Disentangled Recommendation Against User Preference Shifts](https://mlsanity.com/paper/2303.16068/) (2023)
- [Adaptive Hardness Negative Sampling for Collaborative Filtering](https://mlsanity.com/paper/2401.05191/) (2024)
- [On Practical Diversified Recommendation with Controllable Category Diversity Framework](https://mlsanity.com/paper/2402.03801/) (2024)
- [Symmetric Graph Contrastive Learning against Noisy Views for Recommendation](https://mlsanity.com/paper/2408.02691/) (2024)
- [Distributionally Robust Graph Out-of-Distribution Recommendation via Diffusion Model](https://mlsanity.com/paper/2501.15555/) (2025)
- [Hard Negative Sampling via Large Language Models for Recommendation](https://mlsanity.com/paper/2504.04726/) (2025)

## Abstract (arXiv)

Heuristic negative sampling enhances recommendation performance by selecting negative samples of varying hardness levels from predefined candidate pools to guide the model toward learning more accurate decision boundaries. However, our empirical and theoretical analyses reveal that unobserved environmental confounders (e.g., exposure or popularity biases) in candidate pools may cause heuristic sampling methods to introduce false hard negatives (FHNS). These misleading samples can encourage the model to learn spurious correlations induced by such confounders, ultimately compromising its generalization ability under distribution shifts. To address this issue, we propose a novel method named Causal Negative Sampling via Diffusion (CNSDiff). By synthesizing negative samples in the latent space via a conditional diffusion process, CNSDiff avoids the bias introduced by predefined candidate pools and thus reduces the likelihood of generating FHNS. Moreover, it incorporates a causal regularization term to explicitly mitigate the influence of environmental confounders during the negative sampling process, leading to robust negatives that promote out-of-distribution (OOD) generalization. Comprehensive experiments under four representative distribution shift scenarios demonstrate that CNSDiff achieves an average improvement of 13.96% across all evaluation metrics compared to state-of-the-art baselines, verifying its effectiveness and robustness in OOD recommendation tasks.