{"id":"2ab979d3-0448-4023-a60a-926fc34695f0","arxiv_id":"2611.16984","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"ElasticDino uses elastic deformations to upscale DINO ViT features, correcting artifacts and achieving state-of-the-art performance on dense prediction tasks with lower computational cost.","lead":"ElasticDino is a lightweight method that upscales vision transformer features to high resolution while correcting spatial artifacts. It uses elastic deformations guided by the input image to produce sharp, pixel-aligned feature maps efficiently.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader's weakest_assumption focuses on the risk of task-specific semantics leaking into the upsampled features due to training with pseudo-labels. The paper directly addresses this concern: (1) the use of 1x1 convolutional heads prevents spatial mixing, (2) the deformation-based architecture samples from original features rather than editing them, and (3) ablation studies (Table 5, Supplementary Table 3) show that omitting depth or segmentation tasks during training does not degrade performance on those held-out tasks. The reader's concern is valid in principle but is mitigated by the paper's design and evidence. My partial agreement reflects that while the concern is reasonable, the paper provides sufficient counter-evidence. The only potential concern I could identify is whether the deformation fields might introduce folding artifacts (non-invertible mappings), but this is a technical detail that does not threaten the central claim. The paper's own limitation about small objects is acknowledged and analyzed. Therefore, I recommend no change to the ACCEPT verdict.","tokens_in":19514,"tokens_out":527,"duration_ms":7620,"concrete_test":"Verify that the deformation fields are indeed smooth and invertible (or at least not collapsing) by computing the Jacobian determinant of the predicted flow field on a held-out set of 100 images; if more than 5% of pixels have a non-positive determinant, the deformation may introduce folding artifacts that could affect feature semantics.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim is that ElasticDino produces high-resolution, artifact-corrected feature maps that outperform existing upsampling methods on dense prediction tasks. The weakest assumption identified by the reader—that training with 1x1 heads on pseudo-labels might leak task-specific semantics—is explicitly addressed by the authors: they argue that the deformation-based architecture prevents feature-value editing, and they provide ablation studies (Table 5, Supplementary Table 3) showing that held-out tasks (depth, segmentation) remain strong when those tasks are omitted from training. The paper also acknowledges its main limitation (small objects/holes) and provides a size-based analysis (Supplementary Figure 3). The empirical results are extensive, covering multiple backbones (DINOv2, DINOv3, CLIP), multiple datasets, and multiple tasks, with consistent improvements. The computational efficiency claims are supported by Table 1. I find no unresolved load-bearing concern that would undermine the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper introduces ElasticDino, a lightweight adapter that upsamples DINO ViT feature maps while correcting spatial artifacts (positional encoding artifacts and semantic bleeding) through a deformation-based architecture. The method alternates between bilinear upsampling and learned elastic deformations guided by the input image, preserving sharp semantics by sampling from the original features rather than editing feature values. Since no ground-truth high-resolution features exist, training is driven by 1×1 convolutional heads on five dense prediction tasks (RGB, depth, segmentation, normals, intrinsic decomposition) using pseudo-labels from pre-trained models. Extensive experiments on semantic segmentation, depth estimation, and open-vocabulary segmentation across multiple backbones (DINOv2, DINOv3, CLIP) and datasets show consistent improvements over prior upsampling methods (FeatUp, LoftUp, JAFAR, AnyUp, UPLiFT) while being faster and more memory-efficient. The paper also reports ablations confirming that deformations are essential and that held-out tasks remain strong when omitted from training.","tokens_in":19808,"tokens_out":1645,"duration_ms":13462,"significance":"If the results hold, ElasticDino offers a practical and principled solution to a known limitation of ViT-based dense prediction: the low resolution and artifacts of patch-level features. The deformation-based approach is conceptually clean—it avoids feature-value editing and the associated risk of task leakage—and the empirical gains are consistent across multiple backbones, tasks, and datasets. The computational efficiency (Table 1) is a genuine advantage over attention-based competitors. The paper also provides a useful analysis of cross-channel artifact consistency (Section 3.1) that justifies the single-deformation-field design. The main limitation (small objects/holes) is acknowledged and analyzed. Overall, the work represents a solid contribution to the feature-upsampling literature.","major_comments":[{"comment":"The deformation field is defined as F = R(p_f(f) ⊕ p_I(I)) and applied coordinate-wise as f̂_{u,v} = f[F_{u,v}]. The paper states that this 'sampling from the original features' prevents hallucination and preserves semantics. However, the deformation is applied to features at the current resolution, which after the first upsampling step are bilinearly interpolated, not the original low-resolution features. The claim that the model 'diminishes hallucination hazards' by sampling from the original feature map is therefore only strictly true at the first deformation step; at later steps the features have already been interpolated. The authors should clarify what 'original features' means in this context and discuss whether the interpolation step re-introduces any of the smoothness biases the method aims to avoid.","section":"Section 3.2, Eq. (1)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"ElasticDino underperforms on Cityscapes (mIoU 61.33 vs. UPLiFT 65.02, Acc 93.44 vs. 94.34). The authors attribute this to small objects and provide a size-based analysis in Supplementary Figure 3. While this explanation is plausible, the gap is large (≈3.7 mIoU) and Cityscapes is a standard benchmark. The paper should report whether the same size-based pattern holds for the other methods (e.g., does UPLiFT also degrade on small objects, or does it handle them better?). Without this comparison, the claim that the limitation is inherent to the method rather than a weakness of the specific architecture remains partially speculative.","section":"Table 3, Cityscapes row","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The multi-scale loss L_ms sums losses over all stages i, all heads j, and all resolutions r_k. This means gradients flow into every intermediate feature map. The paper states this 'greatly stabilizes and speeds up the learning' but does not ablate this design choice. Since the deformation blocks are already lightweight, it is unclear whether the multi-scale loss is necessary or whether a loss only at the final resolution would suffice. An ablation comparing single-scale vs. multi-scale loss would strengthen the paper.","section":"Section 3.3, multi-scale loss","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The ablation study evaluates depth estimation on a held-out test set from Hypersim. The 'w/o DEFormations' variant shows a catastrophic drop (δ1 from 58.35% to 26.80%). This is a strong result, but the paper does not specify how the 'w/o DEFormations' variant is constructed. Is it simply the same architecture with the deformation blocks removed (i.e., only bilinear upsampling and convolutions)? If so, the comparison conflates the removal of deformations with a change in the number of parameters and the receptive field. The authors should clarify the exact architecture of this ablation and whether the number of parameters was matched.","section":"Section 4.2, Table 5","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The abstract states 'the quadratic complexity of self-attention severely limits feature resolution and induces spatial artifacts.' The quadratic complexity is a computational constraint, not a direct cause of artifacts; the artifacts stem from patchification and positional encoding. Consider rephrasing for precision.","section":"Abstract","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The cross-channel consistency analysis uses Spearman correlation of gradient magnitudes. The reported values (0.29 vs. 0.11 for DINOv2) are modest in absolute terms. The paper should clarify whether these differences are statistically significant (e.g., with confidence intervals or p-values).","section":"Section 3.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The parameter count for ElasticDino is listed as 27,095K (27M), which is substantially larger than competitors (e.g., JAFAR 628K). The paper notes that the model is faster and more memory-efficient despite this, but the large parameter count should be discussed in context: are the extra parameters in the task heads (which are discarded after training) or in the upsampler itself? If the latter, the claim of 'lightweight' should be qualified.","section":"Table 1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The no-depth and no-seg variants show strong performance on held-out tasks, supporting the claim of no task leakage. However, the no-depth variant achieves δ1=60.5% vs. the full model's 64.14%, and the no-seg variant achieves 64.8%. The no-seg variant actually outperforms the full model on depth RMSE (0.56 vs. 0.57). This non-monotonic behavior is interesting and deserves a brief comment.","section":"Supplementary Table 3","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The limitation about small holes is stated but not quantified. The paper should provide a quantitative measure (e.g., performance vs. object size on a standard benchmark) to complement the qualitative inset figure.","section":"Section 5.1","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Reference [6] (Chollet, Xception) is cited for depthwise separable convolutions, which is appropriate. However, the paper uses 'depthwise-separable' and 'depth-wise separable' interchangeably; please standardize.","section":"References","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"minor_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The paper introduces ElasticDino, a lightweight upsampling method for ViT features that uses elastic deformations rather than attention or bilateral filtering. The core insight—that ViT artifacts are cross-channel consistent and can be corrected with a single deformation field—is well supported by a simple correlation analysis (Section 3.1). The architecture is clean: alternating bilinear upsampling and learned deformations, guided by the input image, with no feature-value editing. Training uses 1x1 heads on downstream tasks (segmentation, depth, etc.) to provide supervision without leaking task-specific semantics into the features. The held-out task ablations (Table 5, Supplementary Table 3) back this claim convincingly.\n\nThe empirical work is extensive and honestly reported. The method outperforms prior upsampling approaches (JAFAR, AnyUp, LoftUp, UPLiFT) on most benchmarks, with lower compute and memory (Table 1). The authors also acknowledge the main limitation—poor performance on small objects and holes—and provide a size-based analysis (Supplementary Figure 3) that explains the Cityscapes results. The ablation study is thorough and includes a leave-one-task-out design.\n\nSoft spots: (1) The training relies on pseudo-labels from Mask2Former and DepthAnything. The authors argue this is acceptable because the deformation architecture prevents task leakage, and the held-out experiments support this. I find the argument plausible but not airtight—the pseudo-labels could still bias the deformation field in subtle ways. (2) The method is evaluated only on DINOv2, DINOv3, and CLIP backbones. It would be useful to see results on other ViT families (e.g., MAE, DeiT) to assess generality. (3) The paper claims \"superior accuracy\" on all benchmarks, but Cityscapes is a clear exception (the method is second-best there). The authors address this honestly, but the claim in the abstract is slightly overstated.\n\nOverall, this is a solid, well-written paper with a genuine architectural contribution. The deformation-based approach is novel and the empirical support is strong. I would bring this to a reading group and would cite it. It deserves a serious peer review.\n\nRecommendation: Accept with minor revisions. The authors should add a discussion of pseudo-label bias and ideally test on one additional backbone family.","headline":"ElasticDino is a clean, well-executed paper that introduces a deformation-based upsampling method for ViT features, with strong empirical results and a clear architectural insight.","tokens_in":20251,"tokens_out":579,"duration_ms":7260},"pith_extraction":null,"created_at":"2026-11-13T07:52:35.565427+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":null,"supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}