{"id":"4c7596fd-74e6-49d4-9265-0787753f1d19","arxiv_id":"2605.15196","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"RefDecoder adds reference-image conditioning to video VAE decoders through attention, yielding up to 2.1 dB PSNR gains and better consistency on I2V, editing, and style-transfer tasks.","lead":"The paper proposes RefDecoder, a video VAE decoder that conditions on a reference image by injecting its high-fidelity tokens via attention at each upsampling stage. This targets the asymmetry in current latent diffusion video models where the decoder stays unconditional, leading to lost detail and inconsistency.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Reference attention may fail to preserve structural integrity when denoised latents deviate from the reference, as no fine-tuning or regularization is applied.","rationale":"The reader's weakest assumption directly identifies the load-bearing risk. The architecture description (reference attention at upsampling stages) and the no-fine-tuning claim are plausible but rest on the untested premise that attention alone suffices for all regimes; the reconstruction-to-generation gap is the precise place where this could fail. Full-text details on attention implementation would allow a tighter test, but the proposed check targets the empirical gap without assuming implementation errors.","tokens_in":1752,"tokens_out":372,"duration_ms":38837,"concrete_test":"Take the Wan 2.1 backbone, apply RefDecoder, and recompute VBench I2V scores on 100 generated videos using latents perturbed by additive Gaussian noise (sigma=0.1) before decoding; if subject/background consistency drops below the unconditional baseline or new artifacts appear in >20% of frames, the no-fine-tuning claim weakens.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim requires that co-processing reference tokens (from a lightweight image encoder) with video latents at every decoder upsampling stage via reference attention is sufficient to recover high-fidelity details. This implicitly assumes the attention mechanism can always align and inject the reference signal without creating temporal inconsistencies or new artifacts when the input latent (post-denoising) differs from the reference frame. The reported +2.1 dB PSNR gains are measured on reconstruction benchmarks (Inter4K, WebVid, Large Motion) where latents are clean; the VBench I2V results are cited but without ablations isolating whether gains persist under realistic generation noise or across different decoder backbones without any pipeline adjustment.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript introduces RefDecoder, a reference-conditioned video VAE decoder that injects high-fidelity reference image signals into the decoding process using reference attention at each up-sampling stage. The authors claim this addresses the asymmetry between conditioned denoising networks and unconditional decoders in latent diffusion models for video generation, leading to improved detail preservation and consistency. They report up to +2.1 dB PSNR gains on reconstruction benchmarks like Inter4K, WebVid, and Large Motion, and improvements on VBench I2V without requiring fine-tuning of the pipeline. The method is presented as a plug-and-play module for existing video generation systems and generalizes to tasks like style transfer and video editing.","tokens_in":1906,"tokens_out":599,"duration_ms":37799,"significance":"If the reported gains hold under realistic generation conditions, this could offer a practical plug-in enhancement for video generation by symmetrizing conditioning. The no-fine-tuning claim is a potential strength for adoption. However, the absence of architecture details, training procedures, and ablations on noisy latents (vs. clean reconstruction) makes it difficult to gauge the true significance or robustness of the central architectural change.","major_comments":[{"comment":"Abstract: The central claim that reference attention at decoder upsampling stages is 'sufficient to preserve structural integrity' without fine-tuning or regularization is load-bearing but unsupported by ablations; no experiments test cases where denoised latents deviate from the reference (as occurs in I2V generation), leaving open the risk of temporal inconsistencies or artifacts.","section":"Abstract"},{"comment":"Abstract: Gains of up to +2.1 dB PSNR are reported on reconstruction benchmarks (Inter4K, WebVid, Large Motion) using clean latents, but the VBench I2V results lack controls isolating whether improvements persist under realistic denoising noise or across decoder backbones; this undermines applicability to the stated generation use case.","section":"Abstract"},{"comment":"Abstract: The assertion that RefDecoder 'can be directly swapped into existing video generation systems without additional fine-tuning' requires evidence of integration stability and inference behavior with mismatched latents; none is provided, making the plug-and-play claim unverifiable from the given description.","section":"Abstract"}],"minor_comments":[{"comment":"The abstract references generalization to style transfer and video editing refinement but supplies no quantitative metrics, qualitative examples, or dedicated evaluation sections for these tasks.","section":null}],"recommendation":"major_revision","confidential_remarks":"The provided manuscript text is limited to the abstract with no architecture diagrams, equations, training details, or experimental sections; this raises questions about whether the submission is a full paper or a preliminary note, and whether it aligns with the journal's expectations for technical depth in cs.CV."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the thoughtful comments. We address each major comment below, providing clarifications on our experimental design and claims. We will incorporate additional details into the revised manuscript to address concerns about architecture and training procedures.","responses":[{"response":"We clarify that the reconstruction benchmarks (Inter4K, WebVid, Large Motion) are designed to evaluate the decoder's ability to reconstruct from clean latents, which is the standard protocol for assessing VAE decoders. The VBench I2V results, however, are obtained by integrating RefDecoder into a full latent diffusion pipeline for image-to-video generation, where the input latents are the output of the denoising process and can deviate from the reference. These results show improvements in consistency metrics, indicating that the reference attention helps mitigate inconsistencies even with denoised latents. We did not include separate ablations with artificially corrupted latents, as the end-to-end I2V evaluation serves as the primary validation for the generation use case. We will add a discussion in the revised paper to explicitly distinguish these settings.","revision_made":"partial","referee_comment":"[Abstract] Abstract: The central claim that reference attention at decoder upsampling stages is 'sufficient to preserve structural integrity' without fine-tuning or regularization is load-bearing but unsupported by ablations; no experiments test cases where denoised latents deviate from the reference (as occurs in I2V generation), leaving open the risk of temporal inconsistencies or artifacts."},{"response":"The PSNR gains are indeed reported for clean latent reconstruction to demonstrate the decoder's enhanced fidelity. For the generation use case, the VBench I2V benchmark involves the complete pipeline with realistic denoising noise. We tested RefDecoder on two different decoder backbones (Wan 2.1 and VideoVAE+), reporting consistent improvements in subject consistency, background consistency, and overall quality. To further isolate the effect, we can add more detailed controls in the experiments section of the revision.","revision_made":"partial","referee_comment":"[Abstract] Abstract: Gains of up to +2.1 dB PSNR are reported on reconstruction benchmarks (Inter4K, WebVid, Large Motion) using clean latents, but the VBench I2V results lack controls isolating whether improvements persist under realistic denoising noise or across decoder backbones; this undermines applicability to the stated generation use case."},{"response":"The VBench I2V experiments provide evidence for the plug-and-play nature: RefDecoder is integrated into an existing I2V system (using a pre-trained diffusion model) without any fine-tuning of the diffusion components, and we observe improvements in the metrics. This implies stability during inference with the latents produced by the denoising network, which may be mismatched to the reference in terms of details. We acknowledge that more explicit discussion of the integration process and any potential issues with mismatched latents would be beneficial. We will expand the methods and experiments sections to include architecture details, training procedures, and further analysis of inference behavior.","revision_made":"partial","referee_comment":"[Abstract] Abstract: The assertion that RefDecoder 'can be directly swapped into existing video generation systems without additional fine-tuning' requires evidence of integration stability and inference behavior with mismatched latents; none is provided, making the plug-and-play claim unverifiable from the given description."}],"tokens_in":1508,"tokens_out":718,"duration_ms":57433,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"RefDecoder adds reference attention to video VAE decoder stages and shows measurable PSNR and consistency gains as a drop-in swap. The core move is encoding a reference frame with a lightweight image encoder and mixing those tokens via attention into the decoder at every upsampling stage. This targets the asymmetry where the diffusion denoiser gets heavy conditioning but the decoder stays unconditional. They test the change on backbones like Wan 2.1 and VideoVAE+ and report up to 2.1 dB PSNR lifts on Inter4K, WebVid, and Large Motion reconstruction sets, plus better subject and background consistency on VBench I2V. The no-fine-tuning claim is the practical hook, since it lets people slot the decoder into existing pipelines without touching the rest of the model. It also nods at uses in style transfer and editing refinement, though those get less space. The work is straightforward and the numbers are concrete enough to check. The soft spots sit in the evaluation. Most of the PSNR numbers come from clean reconstruction latents that already align with the reference, so it is not yet clear how the attention behaves when the denoised latent drifts, which is the realistic case in generation. The stress-test note flags exactly this risk of misalignment or new temporal artifacts, and the paper does not appear to include ablations on stage placement, encoder strength, or direct comparisons under noisy latents. Details on the attention implementation and training are also thin in the summary. This is for engineers and researchers who maintain or extend latent video diffusion systems and want a quick decoder upgrade for better input fidelity. A reader already running I2V or editing experiments could try the swap and measure the delta on their own data. It deserves peer review because the method is simple to reproduce and the reported gains are specific, even if referees will likely ask for more tests on actual generation outputs.","headline":"RefDecoder adds reference attention to video VAE decoder stages and shows measurable PSNR and consistency gains as a drop-in swap.","tokens_in":2345,"tokens_out":447,"would_cite":false,"duration_ms":50067,"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":"washburn_uniqueness_aczel","paper_passage":"a reference-conditioned video VAE decoder by injecting high-fidelity reference image signal directly into the decoding process via reference attention"}],"headline":"RefDecoder's reference-attention decoder conditioning is orthogonal to RS recognition-cost forcing","alignment":"orthogonal","rationale":"The paper's core machinery (reference token injection via shared transformer blocks at upsampling stages, latent dropout, L1+LPIPS objective) is a practical CV architecture for VAE decoding. It shares no structural overlap with RS theorems on J-cost uniqueness (Cost.FunctionalEquation.washburn_uniqueness_aczel), φ-ladder constants, 8-tick periodicity, or distinction-to-spacetime forcing (Foundation.RealityFromDistinction). Domain mismatch (video generation vs. parameter-free physics derivation) precludes alignment or contradiction.","tokens_in":56957,"confidence":"high","tokens_out":232,"duration_ms":14855,"cache_read_input_tokens":32896,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-05-15T03:17:21.949834+00:00","model_set":{"reader":"grok-4.3"},"falsifier":null,"supporting_citations":[],"review_version":1}