The Telephone Game: Evaluating Semantic Drift in Unified Models

[2509.04438] The Telephone Game: Evaluating Semantic Drift in Unified Models Skip to main content Search Submit Donate Log in Computer Science > Computer Vision and Pattern Recognition arXiv:2509.04438 (cs) [Submitted on 4 Sep 2025 ( v1 ), last revised 27 Aug 2026 (this version, v3)] Title: The Telephone Game: Evaluating Semantic Drift in Unified Models Authors: Sabbir Mollah , Rohit Gupta , Sirnam Swetha , Qingyang Liu , Ahnaf Munir , Mubarak Shah View PDF HTML (experimental) Abstract: Unified models (UMs) combine visual understanding (I2T) and generation (T2I) in a single framework. We focus on T2I and I2T, where cross-consistency---what a model understands, it should be able to generate---is a promise of unification and a necessity when composing both capabilities. Yet, existing benchmarks evaluate them in isolation: FID/GenEval for T2I; MME/MMBench for I2T. We show this gap is consequential: models scoring competitively on these benchmarks can fail severely when understanding and generation are composed, losing entities, attributes, spatial relations, and counts, resulting in semantic drift. To quantify drift, we introduce the Semantic Drift Protocol (SDP), inspired by the Telephone Game: starting from a caption or image, we alternate I2T and T2I over multiple generations and measure semantic preservation. We propose Mean Cumulative Drift (MCD), an embedding-based measure of content retention across three representation spaces, and Multi-Generation GenEval (MGG), extending GenEval's object-level compliance scoring across generations. To stress-test models beyond COCO-style data, we create a benchmark of 400 image-text pairs sampled from NoCaps and DOCCI, emphasizing novel objects and fine-grained descriptions. Applying SDP to seven models reveals that drift varies dramatically and is not predicted by single-pass scores: BAGEL retains high semantic fidelity over multiple generations, while VILA-U and Janus variants collapse within five generations, despite comparable isolated metrics. We identify six recurring failure modes and find degradation is typically catastrophic rather than gradual: once a critical error occurs, subsequent generations compound it. SDP exposes failure modes that single-pass benchmarks miss, enabling a more faithful assessment of unified model reliability. Code and benchmark: this https URL Subjects: Computer Vision and Pattern Recognition (cs.CV) ; Computation and Language (cs.CL) Cite as: arXiv:2509.04438 [cs.CV] (or arXiv:2509.04438v3 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2509.04438 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Sabbir Mollah [ view email ] [v1] Thu, 4 Sep 2025 17:53:52 UTC (4,271 KB) [v2] Mon, 6 Oct 2025 17:49:39 UTC (19,609 KB) [v3] Thu, 27 Aug 2026 19:48:00 UTC (16,291 KB) Full-text links: Access Paper: View PDF HTML (experimental) TeX Source view license Additional Features Audio Summary Current browse context: cs.CV < prev | next > new | recent | 2025-09 Change to browse by: cs cs.CL References & Citations SciX Google Scholar Semantic Scholar export BibTeX citation Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? ) Connected Papers Toggle Connected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) scite.ai Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle Gotit.pub ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle TXYZ.AI ( What is TXYZ.AI? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators 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 . Which authors of this paper are endorsers? | Disable MathJax ( What is MathJax? ) We gratefully acknowledge support from our major funders , member institutions , and all contributors. About · Help · Contact · Subscribe · Copyright · Privacy · Accessibility · Operational Status (opens in new tab) Major funding support from (opens in new tab) (opens in new tab) (opens in new tab)

Related documents

• The Telephone Game: Evaluating Semantic Drift in Unified Models#684260
• Cross-Entropy Risk Estimation for Language Models: Inconsistency Must Be Dense, and the Holdout Method Is No Exception#616936
• Cross-Entropy Risk Estimation for Language Models: Inconsistency Must Be Dense, and the Holdout Method Is No Exception#616436
• Aligning Thoughts with Answers: Probability Rewards to Tame Thinking Drift#1144150
• Large Language Models Have Unreliable Understanding of Software Engineering Terminology#346584