Pith. sign in

REVIEW 4 major objections 5 minor 30 references

Egocentric Instruction-oriented Affordance Prediction via Large Multimodal Model

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper argues that affordance should be instruction-dependent, and that one large multimodal model can predict it by proposing contact regions and motion directions and then verifying its own proposals in an iterative loop.

desk verdict EIVA is a genuinely useful new task and dataset with a plausible but under-supported motion-direction component; worth refereeing, conditional on releasing the data and answering the egomotion question. read the letter →

arxiv 2508.17922 v1 pith:UDAFWMQR submitted 2025-08-25 cs.RO cs.CV

Bokai Ji , Jie Gu , Xiaokang Ma , Chu Tang , Jingmin Chen , Guangxia Li This is my paper · ORCID
classification cs.ROcs.CV
keywords affordancepredictioninstruction-orientedegocentricvisionlargemultimodalmodelself-refinementrobotmanipulationEIVAdatasetcontactregiongrounding
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that an object's affordance is not a fixed property: the same drawer, door, or kettle offers different contact places and motion directions depending on whether the instruction is to open, close, pick up, or push. To make this claim concrete, it introduces EIVA, a dataset of 14,861 egocentric image-instruction-affordance triplets whose annotations come from an automated pipeline followed by manual review. On this benchmark, the paper shows that large multimodal models can act as instruction-oriented affordance predictors, and that an iterative "search against verifiers" loop, where one LMM proposes an affordance and then verifies and refines its own proposal, outperforms direct prompting and prior affordance methods. If correct, this gives robot manipulation a task-conditioned, open-vocabulary way to decide where and how to act without task-specific training.

What carries the argument

The central object is the "search against verifiers" loop: an Actor LMM generates an affordance proposal, and a Verifier LMM, the same model in a different prompt role, inspects a visualization of that proposal and either approves it or returns corrective feedback, after which the Actor refines and the loop repeats, with a final best-proposal selector. Two visualization mechanisms carry the argument: a SAM-generated mask of the proposed bounding box for the coordinate-based variant, and Set-of-Mark overlays that partition the image into labeled candidate regions for the closed-source model variant. The loop runs in two stages, first fixing the contact region and then fixing a 26-way discrete 3D motion direction, so verification is applied to each decision separately. A second load-bearing mechanism is the EIVA annotation pipeline: peripheral hand-edge points in the contact frame are projected backward through homographies to pre-contact frames to define occlusion-free contact regions, and SpatialTracker's 3D trajectories are summarized by PCA into one principal motion direction.

What would settle it

Independently re-annotate a random sample of EIVA triplets with fresh human labelers who see only the instruction and the pre-contact image, then recompute the reflection-versus-baseline score differences on the corrected labels. If the differences disappear or reverse, the central claim that the Actor-Verifier loop improves instruction-oriented affordance prediction is falsified; if the annotation failure modes appear in a large fraction of samples, the benchmark itself is in question.

Watch

Extended reading notes

Core claim

The central claim is that affordance should be instruction-dependent: different tasks applied to the same object can require different contact regions and different post-contact motion directions, so affordance prediction should take a natural-language instruction as an input rather than reading off static object properties. The paper substantiates this with EIVA, a 14,861-triplet egocentric dataset spanning 146 object-instruction combinations, and with a zero-shot prediction pipeline in which one large multimodal model plays both Actor and Verifier. The Actor proposes a contact region and then a motion direction; the Verifier, prompted with a visualization of the proposal, judges it against the instruction, returns explanations and suggested alternatives when it is wrong, and the loop repeats until approval or a step limit. Across EIVA, the reflection pipeline beats direct LMM prompting and prior affordance methods on contact-region metrics and yields positive, though small, motion-direction cosine similarity; it also generalizes zero-shot to an existing affordance benchmark and improves success rates in robot simulation trials. The paper concludes that the new task, the dataset, and the reflection baseline together open a practical route to instruction-conditioned robot manipulation.

Load-bearing premise

The load-bearing premise is that the automatically generated EIVA labels, contact regions from homography-projected hand-edge points and motion directions from tracked 3D trajectories, really capture the correct place and direction for each instruction, even though the paper documents four annotation failure modes and does not report how often they occur.

Editorial extensions

If this is right

  • An egocentric robot can turn a single LMM into a task-conditioned affordance predictor by prompting it to propose, verify, and refine its own contact regions and motion directions.
  • Instruction-conditioned affordance data of the EIVA kind lets open-vocabulary models outperform methods that treat affordances as static object categories, especially in cluttered real-world scenes.
  • The reflection gains transfer zero-shot to existing affordance benchmarks and to simulated manipulation, so the loop can serve as a usable default baseline for instruction-oriented affordance prediction.
  • Fine-tuning a general-purpose multimodal segmentation model on EIVA unlocks affordance reasoning on unseen object categories, showing the dataset can support training as well as evaluation.
  • Motion-direction prediction is the harder part of the task and remains the main open target, with positive but small cosine-similarity scores.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the verifier's advantage comes mainly from correcting gross localization errors, the same loop could transfer to other dense prediction tasks where LMMs can critique but not generate precise outputs, such as grasp poses or part segmentation.
  • The benchmark's reliability hinges on the unquantified annotation failure modes; measuring their frequency with a fresh human annotation study would tell whether the reported performance gaps are genuine or partly label artifacts.
  • Because the motion-direction scores are small, a natural next step is to replace the 26-direction discretization with continuous or object-relative directions and check whether reflection still adds value.
  • The instruction-dependent view implies that a single image has multiple valid ground-truth affordances for different instructions, so evaluating models against all instructions on the same object, not just the annotated one, would directly measure how well a model disambiguates intent.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper argues that affordance should be instruction-dependent rather than a static object property, and introduces EIVA, a dataset of 14,861 egocentric object-instruction-affordance triplets with contact-region and motion-direction annotations. It also proposes a "search against verifiers" pipeline in which an LMM (Qwen2.5-VL or GPT-4o) iteratively proposes and verifies affordances, and evaluates the approach zero-shot on EIVA, on AGD20K, and in SAPIEN simulations, together with a fine-tuned LISA-13B comparison. The central claim is that the method unlocks new instruction-oriented affordance prediction capabilities and outperforms prior affordance methods on contact-region localization, while motion-direction accuracy is positive but small.

Significance. If the EIVA annotations are reliable, the dataset is a useful benchmark for instruction-conditioned affordance, and the reflection pipeline provides a competitive zero-shot baseline. The paper is commendable for the breadth of its evaluation, including the hand-removal check, the AGD20K generalization test, the fine-tuned LISA comparison, and the transparent documentation of four annotation failure modes. However, the motion-direction evidence is currently too weak to support the paper's central claim that the method 'unlocks' instruction-oriented affordance prediction; the contact-region results are more convincing, but the motion-direction results (CS values of 0.008-0.235) require stronger statistical grounding and a check for egomotion bias.

major comments (4)
  1. [Motion Direction Extraction and Performance Metrics (Appendix)] The ground-truth motion directions are computed as the PCA principal direction of SpatialTracker trajectories in camera space, with no explicit subtraction of egocentric camera motion. Since Ego4D and EPIC-Kitchens are moving-camera videos, the reported directions such as 'backward/up/left' may reflect head/body motion rather than the object's task-relevant displacement. The appendix's failure-mode list does not include this issue, and the reported CS values in Tables 1 and 2 (0.008-0.235) are extremely close to the zero correlation expected from a random discrete direction under the paper's 26-way symmetric discretization. The appendix's statement that 'an acceptable result is that this metric is at least greater than 0' is not a sufficient statistical bar. Please add an egomotion-compensation step or quantify its effect on a validation subset, and report random and majority baselines with confidence intervals and per-category CS breakdowns.
  2. [Correctness of Automatic Annotation (Appendix)] The paper states that all EIVA samples are manually verified to guarantee correctness, but the same appendix identifies four systematic failure modes (wrong active hand side, active hand not visible, wrong contact frame, noisy homography) and does not report how often each occurs or any inter-annotator agreement. Because these failure modes directly corrupt contact-region and motion-direction labels, the benchmark's validity and the conclusion that reflection improves affordance prediction depend on knowing their frequency and whether the manual review actually corrected them. Please report the fraction of samples affected by each failure mode, the review protocol, and inter-annotator agreement on a sample.
  3. [Performance in Robot Simulation Trails (Section and Table 4)] The success rates are reported without trial counts, confidence intervals, or per-instance data; with ten object categories, three instances each, and two tasks per instance, the total of 60 trials leaves very few trials per reported cell, so the 'consistent lift' from reflection is not statistically established. Please report the exact number of trials per method-object cell and per-task, and provide confidence intervals or raw trial outcomes.
  4. [LMM-based Affordance Prediction (Meta Framework) and Table 1] The reflection pipeline is compared against a no-reflection single-proposal baseline, but there is no control that samples the same number of proposals without verifier feedback. The observed gains could partly reflect additional test-time sampling rather than the verifier-informed refinement loop; a baseline that generates T independent proposals and selects one at random (or by a fixed rule) would isolate the contribution of the verifier.
minor comments (5)
  1. [Figure 4] The caption of Figure 4 contains garbled unicode sequences that should be cleaned.
  2. [Section title] 'Robot Simulation Trails' should be 'Robot Simulation Trials'.
  3. [Correctness of Automatic Annotation] The phrase 'relays on the assumption' should be 'relies on the assumption'.
  4. [Details of Finetuning LISA] The text contains the typo 'beacause' and the hyperparameter table lists 'Betas [0.9, 095]', which appears to be a typo for [0.9, 0.95].
  5. [Appendix cross-references] Several cross-references in the appendix are unresolved (e.g., 'in Section') and should be replaced with concrete section numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmark and zero-shot evaluation paper, and its predictions are not constructed from the evaluated labels.

full rationale

This paper does not contain a derivation chain in which a predicted quantity is equivalent to an input by construction. The EIVA dataset labels are produced by an automated pipeline (hand-object detection, SAM2 masks, homography-projected contact points, and PCA over SpatialTracker 3D trajectories) and then manually reviewed; the evaluated LMM predictions are generated zero-shot and compared against those labels, so no parameter or output is fitted to the evaluation set in a way that would make the reported numbers forced. The Actor-Verifier loop uses the same LMM for both roles, which is a genuine methodological self-reference, but it is not a circularity: the claim is empirical, the ablation removing proposal visualization is reported, and independent support is provided by AGD20K comparisons and SAPIEN robot trials. The paper contains no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in through the authors' prior work. The appendix documents annotation failure modes (wrong active hand, occluded active hand, wrong contact frame, noisy homography), but these are concerns about label accuracy and benchmark validity, not about circularity. Similarly, the absence of egomotion compensation in motion-direction extraction and the lack of random/majority baselines for CS are correctness risks, not instances of the paper reducing to its own inputs.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The empirical claims rest on the correctness of automatically generated EIVA labels, on the validity of saliency-style metrics for contact regions, and on zero-shot evaluation assumptions. Several pipeline parameters are unreported or chosen by hand, and the annotation pipeline's failure modes are acknowledged but not quantified.

free parameters (6)
  • Maximum refinement iterations T = 3
    Chosen by hand for both contact-region and motion-direction loops; a larger or smaller T changes the final proposal and therefore all reflection results.
  • Gaussian blur sigma and kernel for affordance maps = not reported
    Applied to projected contact points for ground truth and to predicted masks before metric computation; the scale directly affects SIM, NSS, and AUC-J values.
  • Number of pre-contact frames N = not reported
    Determines how far contact points are projected backward to avoid hand occlusion; the paper does not state its value.
  • Contact point sampling count = not reported
    Number of peripheral hand-mask points used to define the contact region; value is unspecified.
  • DBSCAN parameters for trajectory outlier removal = not reported
    Used to clean SpatialTracker trajectories before PCA; epsilon and min_samples are not given, though trajectory truncation length is stated as 10.
  • Motion direction discretization grid = 26 directions from {-1,0,1}^3 excluding zero
    Design choice that maps continuous PCA directions to discrete labels; coarse discretization can hide direction errors.
assumptions (6)
  • domain assumption Human hand-object contact regions and post-contact 3D motion directions are valid operationalizations of task-oriented affordance.
    EIVA ground truth is defined by these proxy signals; no independent semantic standard is used.
  • domain assumption Timestamped narrations in Ego4D, Epic-Kitchens, and HOI4D correctly label the manipulation intent for each sample.
    Instructions are derived from narrations of the form verb-the-noun; wrong narrations would produce mismatched triplets.
  • domain assumption Homography estimated from background features with hand and object masked accurately projects contact points across frames, and PCA's first eigenvector captures the true motion direction.
    The appendix documents noisy homography, wrong contact frame, and wrong-hand failure modes, so this is a load-bearing modeling assumption.
  • domain assumption The same LMM can serve as a reliable verifier of its own proposals without external ground truth.
    Actor and Verifier share weights; verification feedback is not checked against physics or another model.
  • domain assumption Zero-shot evaluation on EIVA is not inflated by pretraining on the same public video datasets used to build EIVA.
    Ego4D, Epic-Kitchens, and HOI4D are public and may appear in LMM pretraining corpora; the paper does not address contamination.
  • domain assumption Saliency metrics SIM, NSS, and AUC-J are appropriate for scoring contact-region localization.
    These metrics are designed for saliency maps, not fine region masks, and can reward broad or diffuse predictions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Egocentric Instruction-oriented Affordance Prediction via Large Multimodal Model." pith.science (2026). https://pith.science/paper/UDAFWMQR

@misc{pith2026250817922,
  author       = {Pith},
  title        = {Pith review of: Egocentric Instruction-oriented Affordance Prediction via Large Multimodal Model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UDAFWMQR}},
  note         = {Machine review of arXiv:2508.17922}
}
read the original abstract

Affordance is crucial for intelligent robots in the context of object manipulation. In this paper, we argue that affordance should be task-/instruction-dependent, which is overlooked by many previous works. That is, different instructions can lead to different manipulation regions and directions even for the same object. According to this observation, we present a new dataset comprising fifteen thousand object-instruction-affordance triplets. All scenes in the dataset are from an egocentric viewpoint, designed to approximate the perspective of a human-like robot. Furthermore, we investigate how to enable large multimodal models (LMMs) to serve as affordance predictors by implementing a ``search against verifiers'' pipeline. An LMM is asked to progressively predict affordances, with the output at each step being verified by itself during the iterative process, imitating a reasoning process. Experiments show that our method not only unlocks new instruction-oriented affordance prediction capabilities, but also achieves outstanding performance broadly.

Figures

Figures reproduced from arXiv: 2508.17922 by the authors.

Figure 1
Figure 1. Illustration of the data collection process from egocentric videos. Contact points are projected back to pre-contact [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the “search against verifiers” meta framework, which leverages iterative collaboration between an [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison. Colder color indicates higher possibility in the affordance map. Each row corresponds to a [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparison of affordance prediction results with [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the predictions in robot simulation [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Additional qualitative comparison. Each row corresponds to a specific [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Good cases and failure modes of the automatic annotation pipeline. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Average image coverage by object category in Real-World and Laboratory subsets. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Cases of samples before and after eliminating [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Hierarchical distribution of our EIVA15k dataset, [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: Prompting strategy of each step of our affordance prediction pipeline. [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 27 canonical work pages

  1. [1]

    close the fridge

    Language-conditioned affordances : By conditioning affordance predictions on natural language task descrip- tions, our method provides actionable and task-specific guidance, enabling robots to choose different contact regions and manipulation strategies for different goals (e.g., “close the fridge” vs. “open the fridge”)

    work page

  2. [2]

    contact frame

    Real-world generalization : Our model predicts affor- dances directly from real-world visual observations and leverages internet-scale language-vision data. This al- lows for broader applicability beyond simulation and sig- nificantly improves transfer to real-world manipulation scenarios. Moreover, RT-Affordance (Nasiriany et al. 2024) has proposed defin...

    work page 2024

  3. [3]

    2005) measures the correlation between the pre- dicted affordance map and the ground truth

    Normalized Scanpath Saliency (NSS): NSS (Peters et al. 2005) measures the correlation between the pre- dicted affordance map and the ground truth. It is com- puted by normalizing the predicted contact region map to have zero mean and unit standard deviation, followed by averaging over ground-truth locations

    work page 2005

  4. [4]

    Cosine Similarity (CS): CS quantifies the directional similarity between the predicted and ground-truth vec- tors. It is calculated as the cosine of the angle between these two vectors, normalized to a range of [-1, 1], where 1 indicates perfect alignment, 0 indicates orthogonality, and -1 represents complete opposition. An acceptable re- sult is that thi...

    work page

  5. [5]

    During the motion, the position of the grasped part is getting lefter or righter in the image?

    work page

  6. [6]

    Set-of-Mark Prompting Unleashes Extraordinary Vi- sual Grounding in GPT-4V. CoRR. Zhang, Y .; Wu, J.; Li, W.; Li, B.; Ma, Z.; Liu, Z.; and Li, C

    work page

  7. [7]

    grasping

    Video Instruction Tuning With Synthetic Data.CoRR, abs/2410.02713. Preprint. Under review. Justification of Our New Definition of Task-Oriented Affordance While the classical Gibsonian view defines affordance as the action possibilities the environment offers to an agent regardless of the agent’s intentions , this definition, though foundational, does not...

    arXiv

  8. [11]

    Similarity Metric (SIM): SIM (Swain and Ballard

    work page

Show all 30 references
  1. [13]

    2009) is a vari- ant of the AUC metric

    AUC-Judd (AUC-J):AUC-J (Judd et al. 2009) is a vari- ant of the AUC metric. This metric evaluates the pro- portion of the ground truth captured by the predicted af- Preprint. Under review. Model Lid Blender Washing Box gIoU↑ cIoU↑ Sim↑ gIoU↑ cIoU↑ Sim↑ gIoU↑ cIoU↑ Sim↑ gIoU↑ c...

    2009

  2. [16]

    If multiple suitable objects for the task are presented in the image, you should select the one you think the robot is going to manipulate

  3. [17]

    The contact area should be the finest operable part of an object

  4. [18]

    For some doors of appliances, if there isn't a handle, the edge of the door is also considered as a contact area

  5. [19]

    bbox_2d”: [x1, y1, x2, y2], “label

    Real contact area may not be visible in the image because of perspective or occlusion, locate anyway as if it's visible. Examples: - For the task ``Pick up the knife'', the contact area should be the handle of the knife. - For the task ``Open the microwave door'', the contact ...

  6. [20]

    is the contact area marked with color and number [1] suitable for the manipulation task described in the prompt?

  7. [21]

    User {task}

    is the contact area marked in the proposal the finest operable part of the contact area? If multiple suitable contact area are presented in the image, you should select the one you think the robot most likely to manipulate. User {task}. THINK STEP BY STEP:

  8. [22]

    What is the selected contact area (component) in the image?

  9. [23]

    Where should the robot manipulate to finish the task?

  10. [24]

    bbox_2d”: [x1, y1, x2, y2], “label

    Is the selected contact area the proper one to finish the task? Organize your answer in the following format: <THINKING START> [thoughts] <THINKING END> Answer: [Yes/No] If No, please also provide the following information of the contact area that you think is the proper one t...

  11. [25]

    What part of the object is grasped?

  12. [26]

    Imagine the motion of the grasped part as if the task is completed

  13. [27]

    During the motion, the grasped part is getting closer or away from the robot body?

  14. [28]

    During the motion, the position of the grasped part is getting higher or lower in the image?

  15. [30]

    You will be presented with an egocentric image from a robot's perspective, a task description, and a suggestion of the motion direction

    Is the motion direction proposal consistent with the proper motion of the grasped part? Organize your answer in the following format: <THINKING START> [thoughts] <THINKING END> Answer: [Yes/No] If the answer is No, please fill in the following information: - x-axis: [Closer/0/...

  16. [1080]

    Under review

    The high resolution inputs facilitate future evaluation Preprint. Under review. RGB GT Qwen2.5-VL Qwen2.5-VL-Reflection GPT4o-SoM GPT4o-SoM-Reflection OOAL LOCATE VRB 3DOI ManipVQA DoorPotChairBottleBucket OpenCloseOpenClosePick upPick upPullOpenPick upPick upPullDump Figure 6...

    2022

  17. [1991]

    It is computed as the sum of the minimum values at each pixel loca- tion between the predicted contact region map and the ground-truth map

    quantifies the similarity between the predicted af- fordance distribution and the ground truth. It is computed as the sum of the minimum values at each pixel loca- tion between the predicted contact region map and the ground-truth map

  18. [2018]

    Demo2Vec: Reasoning Object Affordances From On- line Videos. In CVPR. Gandhi, K.; Chakravarthy, A.; Singh, A.; Lile, N.; and Good- man, N. D. 2025. Cognitive Behaviors that Enable Self- Improving Reasoners, or, Four Habits of Highly Effective STaRs. arXiv. Geng, Y .; An, B.; G...

    2025

  19. [2021]

    IEEE Trans

    The EPIC-KITCHENS Dataset: Collection, Chal- lenges and Baselines. IEEE Trans. Pattern Anal. Mach. In- tell. Deng, S.; Xu, X.; Wu, C.; Chen, K.; and Jia, K. 2021. 3D AffordanceNet: A Benchmark for Visual Object Affordance Understanding. In CVPR. Do, T.; Nguyen, A.; and Reid, I...

    2021

  20. [2022]

    In NeurIPS

    Flamingo: a Visual Language Model for Few-Shot Learning. In NeurIPS. Anil, R.; Borgeaud, S.; Wu, Y .; Alayrac, J.; Yu, J.; et al

  21. [2023]

    CoRR, abs/2312.11805

    Gemini: A Family of Highly Capable Multimodal Models. CoRR, abs/2312.11805. Bahl, S.; Mendonca, R.; Chen, L.; et al. 2023. Affor- dances from Human Videos as a Versatile Representation for Robotics. In CVPR. Bai, S.; Chen, K.; Liu, X.; Wang, J.; Ge, W.; et al. 2025. Qwen2.5-VL...

    2023 arXiv

  22. [2024]

    CoRR, abs/2410.18558

    Infinity-MM: Scaling Multimodal Performance with Large-Scale and High-Quality Instruction Data. CoRR, abs/2410.18558. Guan, X.; Liu, Y .; Lu, X.; et al. 2024. Search, Verify and Feedback: Towards Next Generation Post-training Paradigm of Foundation Models via Verifier Engineer...

    2024 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.