{"id":"ec848e9f-9dc1-4898-a2bf-a71273156785","arxiv_id":"2505.07322","paper_version":4,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"RealRep uses attribute-disentangled representation learning and a controlled mapping network to generalize SDR-to-HDR conversion across real-world degradations.","lead":"This paper presents RealRep, a framework that learns disentangled luminance and chrominance representations from diverse SDR images to convert them to HDR more robustly than fixed tone mapping. A smart generalist might read it because better SDR-to-HDR conversion could improve viewing quality on modern displays without reshooting content.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Disentanglement's ability to capture unseen SDR distribution shifts lacks direct validation beyond training augmentations","rationale":"The reader's weakest assumption directly names the same condition (disentanglement capturing cross-distribution variations). The concrete test above isolates whether that condition is satisfied outside the authors' augmentation regime, which is the minimal additional evidence needed to move from UNVERDICTED to a higher-confidence verdict.","tokens_in":1697,"tokens_out":342,"duration_ms":31041,"concrete_test":"Collect or synthesize an external test set of SDR images containing degradation types (e.g., HEVC compression at CRF 35 plus Bayer noise) explicitly excluded from the paper's training and validation splits; run the released model and report delta in HDR-VDP-2 and color gamut coverage versus the strongest baseline; if the reported advantage drops below 0.5 dB or 5% gamut coverage, the disentanglement-guided adaptation does not hold for the claimed robustness.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that Realistic Attribute-Disentangled Representation Learning plus Luma-/Chroma-aware negative exemplar generation produces priors that generalize across arbitrary real-world SDR degradations. The abstract describes constructing degradation-sensitive contrastive pairs, but this construction is internal to the training distribution; if the generated negatives only span a narrow set of simulated tone and noise variations, the attribute-level priors fed to DDACMNet's control-aware normalization may fail to modulate correctly on out-of-distribution SDR inputs (e.g., heavy compression artifacts or sensor-specific gamut clipping absent from the augmentation pipeline). This is the least secure link for the generalization half of the headline claim.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes RealRep, a generalized SDR-to-HDR conversion framework. It introduces Realistic Attribute-Disentangled Representation Learning to explicitly disentangle luminance and chrominance components for capturing intrinsic content variations across SDR distributions, a Luma-/Chroma-aware negative exemplar generation strategy to construct degradation-sensitive contrastive pairs, and the Degradation-Domain Aware Controlled Mapping Network (DDACMNet), a lightweight two-stage architecture that performs adaptive hierarchical mapping via control-aware normalization conditioned on degradation features. The central claim is that this approach yields superior generalization and perceptually faithful HDR color gamut reconstruction compared to state-of-the-art methods, as supported by extensive experiments.","tokens_in":1833,"tokens_out":520,"duration_ms":25693,"significance":"If the generalization results hold under rigorous out-of-distribution testing, the work would be significant for computer vision and multimedia processing by moving beyond fixed tone-mapping operators to handle diverse real-world SDR degradations. The explicit disentanglement of attributes combined with contrastive modeling of tone discrepancies offers a principled way to build degradation-aware priors, which could improve robustness in practical HDR-WCG pipelines. The lightweight design of DDACMNet is a practical strength if the performance gains are reproducible.","major_comments":[{"comment":"The generalization claim (abstract and §4) rests on the assumption that attribute-disentangled priors from internally generated contrastive pairs transfer to arbitrary real-world SDR inputs. However, the Luma-/Chroma-aware negative exemplar generation is described as spanning simulated tone and noise variations within the training distribution; without explicit evaluation on out-of-distribution cases such as heavy compression artifacts or sensor-specific gamut clipping, the robustness of DDACMNet's control-aware normalization remains unverified and load-bearing for the headline result.","section":"Abstract and Experiments (§4)"}],"minor_comments":[{"comment":"Clarify the exact formulation of the contrastive loss and how the degradation-conditioned features are injected into the normalization layers of DDACMNet; the current description leaves the control mechanism somewhat underspecified.","section":"Method (§3)"},{"comment":"Ensure all quantitative tables report both mean and standard deviation across multiple runs or datasets to allow assessment of statistical significance of the reported outperformance.","section":"Experiments (§4)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript appears to fit the scope of a computer vision journal focused on image processing and enhancement. The citation pattern is standard for the subfield; no obvious issues with novelty disclosure."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback on our generalization claims. We address the major comment point-by-point below and will incorporate revisions to strengthen the evaluation of out-of-distribution robustness.","responses":[{"response":"We appreciate the referee's careful reading. The Luma-/Chroma-aware negative exemplar generation indeed relies on simulated tone and noise variations to construct contrastive pairs during training. However, the training data itself comprises diverse real-world SDR sources with natural degradations, and our test sets include real inputs exhibiting compression artifacts, noise, and gamut variations. The DDACMNet's control-aware normalization is explicitly conditioned on degradation features extracted from these real inputs, which enables the observed generalization. That said, we acknowledge that dedicated, isolated OOD benchmarks for heavy compression and sensor-specific gamut clipping are not separately reported. In the revision we will add such targeted evaluations (including quantitative metrics and qualitative results on held-out degradation types) to directly verify the robustness of the control mechanism.","revision_made":"yes","referee_comment":"[Abstract and Experiments (§4)] The generalization claim (abstract and §4) rests on the assumption that attribute-disentangled priors from internally generated contrastive pairs transfer to arbitrary real-world SDR inputs. However, the Luma-/Chroma-aware negative exemplar generation is described as spanning simulated tone and noise variations within the training distribution; without explicit evaluation on out-of-distribution cases such as heavy compression artifacts or sensor-specific gamut clipping, the robustness of DDACMNet's control-aware normalization remains unverified and load-bearing for the headline result."}],"tokens_in":1381,"tokens_out":338,"duration_ms":29362,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main point is that this paper builds a framework called RealRep to convert SDR to HDR by explicitly separating luminance and chrominance attributes, then feeding those into a degradation-aware network for adaptive mapping. That combination is the actual new piece compared to the fixed operators they cite as baselines. They add luma- and chroma-aware negative exemplar generation to create contrastive pairs that model tone differences across SDR styles, and wrap it in DDACMNet, a lightweight two-stage setup that uses control-aware normalization to modulate the output based on detected degradations. This setup directly tackles the practical issue of varied real-world SDR content like compression artifacts or gamut shifts, which fixed methods handle poorly. The abstract frames the experiments as showing consistent gains in generalization and perceptual fidelity, and the approach feels grounded in addressing distribution shifts rather than just fitting more parameters. On the soft spots, the stress-test concern about whether the generated negatives span enough unseen degradations holds some weight here. If the augmentations stay mostly within simulated tone and noise variations, the attribute priors might not steer the mapping correctly on inputs with heavy sensor clipping or compression not seen in training. That could weaken the robustness half of the claim, even if the in-distribution results look fine. The paper does not appear to lean on circular definitions or unfalsifiable fitting, which helps. This work is aimed at computer vision researchers and media production teams dealing with HDR display pipelines. A reader focused on practical conversion tools would find the framework and the specific negative sampling strategy useful to build on. It deserves a serious referee because the technical choices are explicit and the problem is timely, even if the experiments will likely need more out-of-distribution testing and clearer ablation breakdowns during review.","headline":"RealRep adds a disentangled luma-chroma approach with controlled mapping that targets real SDR degradations better than fixed tone operators, though the generalization edge still needs tighter validation.","tokens_in":2332,"tokens_out":419,"would_cite":false,"duration_ms":23019,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[{"relation":"unclear","rs_module":"IndisputableMonolith/Cost/FunctionalEquation.lean","rs_theorem":null,"paper_passage":"RealRep, which explicitly disentangles luminance and chrominance to reduce style variation and builds a robust embedding space for accurate SDR-to-HDR conversion."}],"headline":"Standard CV disentanglement pipeline with no RS-shaped cost or distinction machinery","alignment":"orthogonal","rationale":"The paper's core is a contrastive multi-view encoder (luma/chroma global/local) plus degradation-conditioned affine modulation in DDACMNet. This is conventional representation learning for image restoration; it never invokes a reciprocal cost J, golden-ratio fixed points, 8-tick periodicity, or any parameter-free derivation from a single distinction. No theorem in the RS corpus (e.g., reality_from_one_distinction, washburn_uniqueness_aczel, or Alexander-duality D=3 forcing) is paralleled or contradicted.","tokens_in":51019,"confidence":"high","tokens_out":227,"duration_ms":11088,"cache_read_input_tokens":32896,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"RealRep uses attribute-disentangled learning to generalize SDR-to-HDR conversion across real-world degradations.","keywords":["SDR-to-HDR conversion","attribute disentanglement","representation learning","tone mapping","color gamut","degradation awareness","contrastive learning","adaptive mapping"],"falsifier":"A counterexample would be a set of real-world SDR images with novel degradation types where the method produces color shifts or loses detail in HDR output compared to baselines.","tokens_in":2605,"feed_emoji":"","tokens_out":584,"duration_ms":29432,"temperature":0.7,"pith_summary":"The paper proposes RealRep, a framework for converting Standard Dynamic Range images to High Dynamic Range that learns to separate luminance and chrominance attributes. This separation helps capture how different SDR content varies in appearance and quality. By generating contrastive pairs sensitive to these differences and using a controlled mapping network, the method adapts the conversion process to various degradations. The result is more consistent and accurate HDR reconstructions than previous fixed approaches. This matters because real SDR content comes from many sources with inconsistent quality, making a one-size-fits-all tone mapping insufficient.","feed_headline":"Disentangled learning improves real-world SDR-to-HDR conversion","feed_subtitle":"By separating luminance and chrominance, the approach adapts to varied degradations for more faithful HDR results.","key_machinery":"Realistic Attribute-Disentangled Representation Learning (RealRep) that disentangles luminance and chrominance to model intrinsic content variations, paired with the Degradation-Domain Aware Controlled Mapping Network (DDACMNet) that uses control-aware normalization for adaptive mapping.","core_discovery":"The central discovery is that explicitly disentangling luminance and chrominance components through Realistic Attribute-Disentangled Representation Learning, combined with luma-chroma aware negative exemplars and a degradation-domain aware controlled mapping network, enables robust adaptive hierarchical mapping from diverse SDR inputs to perceptually faithful HDR outputs with wide color gamut.","pith_inferences":["Similar disentanglement could extend to video SDR-to-HDR for temporal consistency.","Applying this to other image enhancement tasks like low-light or super-resolution might benefit from attribute separation.","Testing on synthetic degradations could validate the robustness claims further."],"forward_implications":["Adaptive mapping handles diverse SDR styles and degradations better than fixed operators.","Contrastive learning with degradation-sensitive pairs improves modeling of tone discrepancies.","Two-stage framework allows hierarchical adaptation guided by degradation features.","Improved generalization leads to perceptually faithful color gamut reconstruction across distributions."],"fun_headline_variants":["RealRep disentangles attributes for SDR-to-HDR conversion","Disentangling luminance from chrominance in SDR-to-HDR","Luma-chroma aware negative exemplars for SDR-to-HDR","Degradation aware controlled mapping network for SDR-to-HDR"],"cache_read_input_tokens":64,"weakest_assumption_plain":"Disentangling luminance and chrominance components captures enough intrinsic variations in SDR content to enable reliable adaptive mapping.","fun_headline_variants_meta":{"raw":{"variants":["RealRep disentangles attributes for SDR-to-HDR conversion","Disentangling luminance from chrominance in SDR-to-HDR","Luma-chroma aware negative exemplars for SDR-to-HDR","Degradation aware controlled mapping network for SDR-to-HDR"]},"model":"grok-4.3","cost_usd":0.014882,"raw_usage":{"total_tokens":6388,"prompt_tokens":655,"num_sources_used":0,"completion_tokens":68,"cost_in_usd_ticks":148824500,"prompt_tokens_details":{"text_tokens":655,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":5665,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":655,"tokens_out":68,"duration_ms":60085,"temperature":1.0,"reasoning_tokens":5665,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-22T15:50:48.229692+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A counterexample would be a set of real-world SDR images with novel degradation types where the method produces color shifts or loses detail in HDR output compared to baselines.","supporting_citations":[],"review_version":1}