aiqbench analysis
This research addresses a critical failure mode in Retrieval-Augmented Generation (RAG) systems where retrieved context contradicts a model's internal parametric knowledge. The authors propose the Trajectory Variance Score (TVS), a lightweight metric that detects these conflicts by analyzing semantic divergence across parallel stochastic denoising trajectories in discrete diffusion models. By achieving strong detection accuracy with minimal inference overhead, this work offers a practical method for monitoring and mitigating hallucinations caused by conflicting information sources in advanced generative architectures.
The Temporal Tug-of-War: Visualizing and Detecting RAG Conflicts in Diffusion Models via Trajectory Variance
arXiv:2609.31684v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) introduces a specific failure mode in discrete diffusion language models: when retrieved context contradicts parametric knowledge, the iterative denoising process becomes a visible battleground between competing knowledge sources. We identify temporal semantic divergence as an observable for detecting these conflicts and introduce the Trajectory Variance Score (TVS), a simple and interpretable measure of this divergence. TVS computes the mean pairwise cosine distance of answer embeddings across independent stochastic denoising trajectories, capturing the temporal tug of war between parametric and contextual attractors. Requiring as few as two parallel inference runs, TVS is computationally lightweight. Across four diverse datasets (Synthetic, SciQ, PopQA, and CounterFact), a simple Logistic Regression classifier using TVS achieves $70.10\%$ accuracy and $0.7647$ AUROC on LLaDA. On Dream 7B, increasing the number of trajectories from two to five improves accuracy from $63.91\%$ to $69.62\%$. More complex sequential models provide only marginal improvements over the linear classifier. Evaluation across LLaDA and Dream 7B demonstrates that conflict-induced trajectory dynamics and their key properties transfer across distinct diffusion architectures.
Abstract:Retrieval-Augmented Generation (RAG) introduces a specific failure mode in discrete diffusion language models: when retrieved context contradicts parametric knowledge, the iterative denoising process becomes a visible battleground between competing knowledge sources. We identify temporal semantic divergence as an observable for detecting these conflicts and introduce the Trajectory Variance Score (TVS), a simple and interpretable measure of this divergence. TVS computes the mean pairwise cosine distance of answer embeddings across independent stochastic denoising trajectories, capturing the temporal tug of war between parametric and contextual attractors. Requiring as few as two parallel inference runs, TVS is computationally lightweight. Across four diverse datasets (Synthetic, SciQ, PopQA, and CounterFact), a simple Logistic Regression classifier using TVS achieves $70.10\%$ accuracy and $0.7647$ AUROC on LLaDA. On Dream 7B, increasing the number of trajectories from two to five improves accuracy from $63.91\%$ to $69.62\%$. More complex sequential models provide only marginal improvements over the linear classifier. Evaluation across LLaDA and Dream 7B demonstrates that conflict-induced trajectory dynamics and their key properties transfer across distinct diffusion architectures.
Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2609.31684 [cs.CL]
(or arXiv:2609.31684v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.31684 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 266 words
Related coverage
Uncertainty-Aware Conflict Detection Against Operator-Conditioned Weather Hazards
arXiv · Sep 14, 2026
MLLM-Guided Semantic Correction for Text-to-Video Generation
arXiv · Aug 18, 2026
Robots as Tokens: Unified Diffusion Transformer for Coordinated Multi-Robot Trajectory Generation
arXiv · Jun 16, 2026
Automating coherent long-form video generation
Google Research · Sep 24, 2026