Paper deep dive
A Calibration Audit of Confidence in Feed-Forward 3D Reconstruction
Nanxing Nick Deng, Qing Cheng, Niclas Zeller, Daniel Cremers
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 94%
Last extracted: 9/3/2026, 2:52:18 AM
Summary
This paper audits the calibration of confidence scores in feed-forward 3D reconstruction models. It finds that while confidence ranks error well, it is systematically overconfident (off by 2.4x on average) when applied to unseen data. The authors propose a power-law rescaling to correct magnitude but note that scene-specific shape errors remain uncorrected. The paper was subsequently withdrawn.
Entities (7)
Relation Signals (5)
Qing Cheng → withdrew → arXiv:2608.29705
confidence 99% · This paper has been withdrawn by Qing Cheng
per-pixel confidence → exhibits → overconfidence
confidence 95% · The median case is off by 2.4x across all seven models, and the error prediction is further off the more confident the model is.
per-pixel confidence → istrainedas → loss weight
confidence 95% · It is trained as a loss weight, not as an uncertainty magnitude
Feed-forward 3D reconstruction models → outputs → per-pixel confidence
confidence 95% · Feed-forward 3D reconstruction models emit a per-pixel confidence that downstream systems read as a reliability signal.
power law → corrects → predicted uncertainty magnitude
confidence 90% · A power law with two constants per backbone and dataset corrects the overall magnitude of the predicted uncertainty and leaves the ranking untouched.
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:Feed-forward 3D reconstruction models emit a per-pixel confidence that downstream systems read as a reliability signal. It is trained as a loss weight, not as an uncertainty magnitude, and whether it can be used as an error prediction has not been measured. We audit seven released backbones on thirteen datasets and score the confidence on four properties, how well it ranks error, whether its level is right on average, whether it holds across the confidence range, and whether its intervals cover the truth. The confidence ranks error well, but the predicted uncertainty is too low when it is read under conditions that are not exactly those of training. The median case is off by 2.4x across all seven models, and the error prediction is further off the more confident the model is. We show that this phenomenon can appear even though the loss's optimum is reached. A released model resumed under its own loss reaches that optimum on its training data within a few hundred updates and stays overconfident on unseen frames. A power law with two constants per backbone and dataset corrects the overall magnitude of the predicted uncertainty and leaves the ranking untouched. What no rescaling reaches is the scene, which we attribute to the model's missing knowledge of scale across predictions. Every correction we tried is close to right on average and still leaves two thirds of held-out scenes outside a five-point band, because what a scene is missing is a shape rather than a shift. We release the audit protocol, its results, and the fitted constants per model and dataset. Fitted with the target dataset held out, the constants bring the median case from 2.4x off to 1.35x, and a refit on a few labelled scenes of that dataset reaches 1.12x.
Tags
Links
- Source: https://arxiv.org/abs/2608.29705v2
- Canonical: https://arxiv.org/abs/2608.29705v2
PDF not stored locally. Use the link above to view on the source site.
Full Text
5,257 characters extracted from source content.
Expand or collapse full text
Skip to main content Search Submit Donate Log in Search arXiv Press Enter to search · Advanced search Computer Science > Computer Vision and Pattern Recognition arXiv:2608.29705v2 (cs) This paper has been withdrawn by Qing Cheng [Submitted on 30 Aug 2026 (v1), last revised 2 Sep 2026 (this version, v2)] Title:A Calibration Audit of Confidence in Feed-Forward 3D Reconstruction Authors:Nanxing Nick Deng, Qing Cheng, Niclas Zeller, Daniel Cremers View a PDF of the paper titled A Calibration Audit of Confidence in Feed-Forward 3D Reconstruction, by Nanxing Nick Deng and 3 other authors No PDF available, click to view other formats Abstract:Feed-forward 3D reconstruction models emit a per-pixel confidence that downstream systems read as a reliability signal. It is trained as a loss weight, not as an uncertainty magnitude, and whether it can be used as an error prediction has not been measured. We audit seven released backbones on thirteen datasets and score the confidence on four properties, how well it ranks error, whether its level is right on average, whether it holds across the confidence range, and whether its intervals cover the truth. The confidence ranks error well, but the predicted uncertainty is too low when it is read under conditions that are not exactly those of training. The median case is off by 2.4x across all seven models, and the error prediction is further off the more confident the model is. We show that this phenomenon can appear even though the loss's optimum is reached. A released model resumed under its own loss reaches that optimum on its training data within a few hundred updates and stays overconfident on unseen frames. A power law with two constants per backbone and dataset corrects the overall magnitude of the predicted uncertainty and leaves the ranking untouched. What no rescaling reaches is the scene, which we attribute to the model's missing knowledge of scale across predictions. Every correction we tried is close to right on average and still leaves two thirds of held-out scenes outside a five-point band, because what a scene is missing is a shape rather than a shift. We release the audit protocol, its results, and the fitted constants per model and dataset. Fitted with the target dataset held out, the constants bring the median case from 2.4x off to 1.35x, and a refit on a few labelled scenes of that dataset reaches 1.12x. Comments: Need to improve the writing Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) Cite as: arXiv:2608.29705 [cs.CV] (or arXiv:2608.29705v2 [cs.CV] for this version) https://doi.org/10.48550/arXiv.2608.29705 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Qing Cheng [view email] [v1] Sun, 30 Aug 2026 10:21:34 UTC (808 KB) [v2] Wed, 2 Sep 2026 14:04:24 UTC (1 KB) (withdrawn) Full-text links: Access Paper: View a PDF of the paper titled A Calibration Audit of Confidence in Feed-Forward 3D Reconstruction, by Nanxing Nick Deng and 3 other authorsWithdrawn No license for this version due to withdrawn Current browse context: cs.CV < prev | next > new | recent | 2026-08 Change to browse by: cs cs.AI References & Citations NASA ADSGoogle Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: 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?) mathjaxToggle(); 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