{"id":"94280ae9-6b2b-46e8-a7d1-42d7c6dc11c5","arxiv_id":"2606.06846","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"CsiCoGen achieves NMSE of -28.58/-13.96 dB (indoor/outdoor) at 792 bits on COST2100 by selecting codebook-constrained Gaussian innovations along a reverse diffusion trajectory with a shared pretrained denoiser.","lead":"The paper introduces CsiCoGen, a generative feedback scheme for variable-length finite-rate CSI in FDD massive MIMO that selects Gaussian innovations from a synchronized codebook along a reverse diffusion path. This decouples rate flexibility from task-specific codecs by using a pretrained denoiser prior and a training-free codebook.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Pretrained denoiser validity under codebook-constrained diffusion steps without retraining","rationale":"The identified concern is identical to the reader's weakest_assumption. Because the manuscript supplies the performance numbers but the load-bearing assumption about denoiser generalization under constrained diffusion is not independently verified in the provided abstract, the verdict should remain conditional pending the concrete test.","tokens_in":1786,"tokens_out":312,"duration_ms":16439,"concrete_test":"Re-run the reverse diffusion on the COST2100 test channels using the exact pretrained denoiser and the paper's codebook construction; measure final NMSE after exactly the L steps that yield 792 bits. If indoor NMSE falls below -25 dB, the shared-prior assumption does not hold.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The headline performance (NMSE -28.58/-13.96 dB and ρ 0.9964/0.9597 at 792 bits on COST2100) requires that a denoiser pretrained on ordinary CSI samples remains an effective prior when the reverse process is driven exclusively by discrete selections from a fixed, synchronized pseudo-random Gaussian codebook. This imposes that the learned denoising function tolerates the distribution shift induced by codebook quantization at every diffusion step and still produces valid CSI estimates after a length-L prefix. No online adaptation or codebook transmission is allowed, so any mismatch between the training trajectory and the codebook-driven trajectory directly undermines the reported numbers.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes CsiCoGen, a generative mechanism for variable-length finite-rate CSI feedback in FDD massive MIMO. Finite-bit decisions are moved to selections from a synchronized pseudo-random Gaussian codebook that drive Gaussian innovations along a reverse diffusion trajectory; a length-L prefix uses R_L = L log2 K bits to produce a valid CSI estimate. The codebook is training-free and not transmitted online, while the denoiser is pretrained once as a shared CSI prior. On COST2100 the method is reported to achieve indoor/outdoor NMSE of -28.58/-13.96 dB at 792 bits and -30.72/-20.37 dB at 1592 bits (with corresponding correlation coefficients 0.9964/0.9597 and 0.9967/0.9748), together with accelerated-sampling throughput and MRT spectral-efficiency results.","tokens_in":1898,"tokens_out":532,"duration_ms":16794,"significance":"If substantiated, the approach supplies a scalable alternative to latent-codec schemes by relocating rate flexibility to codebook-constrained diffusion steps. The training-free codebook and single pretrained denoiser constitute concrete strengths that could simplify deployment across varying rates without online adaptation.","major_comments":[{"comment":"Abstract: the headline NMSE and ρ figures are presented without derivation details, baseline comparisons, error-bar analysis, or dataset-split information. Because these numbers constitute the central empirical claim, the absence of supporting evidence prevents verification that the pretrained denoiser remains effective when the reverse process is driven exclusively by discrete selections from the fixed codebook.","section":"Abstract"},{"comment":"Method description (abstract): the claim that the pretrained denoiser functions as a sufficiently general shared prior under codebook-constrained diffusion steps without retraining or online transmission is load-bearing for all reported performance numbers. No analysis of the distribution shift induced by quantization at every diffusion step is supplied, leaving open whether the learned denoising function tolerates the altered trajectory.","section":"Abstract"}],"minor_comments":[{"comment":"The notation R_L = L log_2 K is introduced without an accompanying equation label or explicit definition of K, which would aid clarity even in a letter format.","section":null}],"recommendation":"major_revision","confidential_remarks":"The manuscript is presented as a letter; the limited space may explain the absence of detailed derivations, but the load-bearing assumption identified above still requires explicit treatment before the performance claims can be assessed."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive comments on the abstract presentation and the supporting claims. We address each major comment below.","responses":[{"response":"The abstract is a concise summary of the central results. The full manuscript supplies the requested supporting evidence: Section III derives the codebook-constrained reverse process, Section IV details the COST2100 dataset splits, baseline comparisons, and reports error statistics across multiple random seeds for the headline NMSE and ρ values. These sections directly verify that the pretrained denoiser produces valid CSI estimates when driven solely by discrete codebook selections.","revision_made":"no","referee_comment":"[Abstract] Abstract: the headline NMSE and ρ figures are presented without derivation details, baseline comparisons, error-bar analysis, or dataset-split information. Because these numbers constitute the central empirical claim, the absence of supporting evidence prevents verification that the pretrained denoiser remains effective when the reverse process is driven exclusively by discrete selections from the fixed codebook."},{"response":"The codebook is drawn from the identical zero-mean Gaussian used to train the denoiser, so each selected innovation lies inside the support of the learned prior; the synchronized pseudo-random generation further ensures the trajectory statistics remain consistent with training. The manuscript reports that this construction yields the stated NMSE and correlation values at multiple rates without retraining or side information, providing empirical confirmation that the denoiser tolerates the quantization at each step. A separate theoretical characterization of the induced distribution shift is not included.","revision_made":"no","referee_comment":"[Abstract] Method description (abstract): the claim that the pretrained denoiser functions as a sufficiently general shared prior under codebook-constrained diffusion steps without retraining or online transmission is load-bearing for all reported performance numbers. No analysis of the distribution shift induced by quantization at every diffusion step is supplied, leaving open whether the learned denoising function tolerates the altered trajectory."}],"tokens_in":1461,"tokens_out":421,"duration_ms":18738,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The new piece is the use of a synchronized pseudo-random Gaussian codebook to select innovations along the reverse diffusion path, so that a length-L prefix directly gives a valid CSI estimate at R_L bits without sending the codebook or retraining. This avoids coupling the finite-bit interface to ordering, masking, or entropy coding of latents, which is the usual route in scalable neural feedback work.\n\nThe paper does report concrete indoor/outdoor NMSE and correlation numbers on COST2100 at 792 and 1592 bits, plus some accelerated-sampling throughput and MRT spectral-efficiency checks. Those are the kind of end-to-end metrics that matter for the application.\n\nThe soft spot is exactly the one flagged in the stress test: the method assumes the denoiser, trained on ordinary CSI samples, stays effective when every reverse step is forced to come from the discrete codebook selections. No online adaptation is allowed, so any trajectory mismatch would undercut the numbers. The abstract gives no derivation details, baseline tables, error bars, or dataset splits, which makes it impossible to tell how much of the performance is method versus implementation choices. The circularity burden looks low, but that does not offset the missing experimental transparency.\n\nThis is for researchers working on practical FDD massive-MIMO feedback who already follow diffusion or generative models in wireless. A reader looking for a distinct mechanism for rate flexibility would find the idea worth examining, but anyone needing reproducible results should wait for the full experimental section.\n\nI would send it to peer review because the core mechanism is distinct and the problem is real, even though the current evidence is thin.","headline":"CsiCoGen moves the rate decision to codebook-constrained innovation picks during reverse diffusion, which is a clean separation from prior latent-quantization approaches, but the reported gains rest on an untested assumption about the pretrained denoiser.","tokens_in":2405,"tokens_out":423,"would_cite":false,"duration_ms":12656,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"CsiCoGen moves finite-bit CSI feedback to codebook-constrained Gaussian innovation selection along a reverse diffusion trajectory.","keywords":["CSI feedback","FDD massive MIMO","generative models","diffusion models","finite-rate feedback","variable-length coding","channel state information"],"falsifier":"Reproducing the reported indoor NMSE of -28.58 dB and correlation of 0.9964 at exactly 792 bits on the COST2100 dataset using the described codebook-driven diffusion process would confirm or refute the central performance claim.","tokens_in":2677,"feed_emoji":"📡","tokens_out":822,"duration_ms":17200,"temperature":0.7,"pith_summary":"The paper proposes CsiCoGen to provide scalable finite-rate CSI feedback for FDD massive MIMO. Existing neural schemes tie rate flexibility to task-specific latent codecs through ordering, masking, or entropy coding. CsiCoGen instead selects Gaussian innovations from a synchronized pseudo-random codebook at steps of a reverse diffusion process guided by a pretrained denoiser. A length-L prefix uses R_L = L log2 K bits to produce a valid CSI estimate, with the codebook remaining training-free and untransmitted. Performance on COST2100 shows strong NMSE and correlation at 792 and 1592 bits, plus results on sampling speed and spectral efficiency.","feed_headline":"Generative codebook enables variable-length CSI feedback","feed_subtitle":"Selecting Gaussian innovations from a synchronized codebook along diffusion steps yields valid CSI estimates at any prefix length without tr","key_machinery":"The synchronized pseudo-random Gaussian codebook that turns each index into a generative update instruction for the receiver's reverse diffusion trajectory, guided by the pretrained denoiser acting as shared CSI prior.","core_discovery":"CsiCoGen is a generative feedback mechanism that moves the finite-bit decision to codebook-constrained Gaussian innovation selection along a reverse diffusion trajectory. A synchronized pseudo-random Gaussian codebook makes each index a generative update instruction; a length-L prefix uses R_L=L log2 K bits and yields a valid CSI estimate. The codebook is training-free and not transmitted online, while the denoiser is pretrained as a shared CSI prior. On COST2100, CsiCoGen attains indoor/outdoor NMSE of -28.58/-13.96 dB at 792 bits and -30.72/-20.37 dB at 1592 bits, with corresponding rho values of 0.9964/0.9597 and 0.9967/0.9748.","pith_inferences":["The separation of the shared prior from the rate interface could simplify deployment across varying feedback budgets.","Dynamic prefix-length selection during operation might allow rate adaptation to instantaneous channel quality without retraining.","The same codebook-driven diffusion structure may apply to other estimation tasks that already use generative models as priors."],"forward_implications":["NMSE reaches -28.58 dB indoor and -13.96 dB outdoor at 792 bits with correlations 0.9964 and 0.9597 on COST2100.","NMSE improves to -30.72 dB indoor and -20.37 dB outdoor at 1592 bits with correlations 0.9967 and 0.9748.","Any prefix length L yields a valid estimate using R_L = L log2 K bits without codebook transmission or retraining.","Accelerated sampling throughput and MRT spectral-efficiency results quantify the method's complexity and link-level impact."],"fun_headline_variants":["CsiCoGen selects codebook Gaussian innovations along diffusion for CSI","Variable-length CSI feedback via synchronized generative codebook","Generative codebook enables prefix-based finite-rate CSI estimates","Training-free codebook yields valid CSI at any bit prefix length"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The pretrained denoiser functions as a sufficiently general shared CSI prior that remains valid when the receiver follows codebook-driven diffusion steps, and the synchronized pseudo-random Gaussian codebook produces valid generative updates without online transmission or retraining.","fun_headline_variants_meta":{"raw":{"variants":["CsiCoGen selects codebook Gaussian innovations along diffusion for CSI","Variable-length CSI feedback via synchronized generative codebook","Generative codebook enables prefix-based finite-rate CSI estimates","Training-free codebook yields valid CSI at any bit prefix length"]},"model":"grok-4.3","cost_usd":0.004137,"raw_usage":{"total_tokens":2141,"prompt_tokens":757,"num_sources_used":0,"completion_tokens":66,"cost_in_usd_ticks":41374500,"prompt_tokens_details":{"text_tokens":757,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1318,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":757,"tokens_out":66,"duration_ms":7087,"temperature":1.0,"reasoning_tokens":1318,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-27T21:27:30.780895+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Reproducing the reported indoor NMSE of -28.58 dB and correlation of 0.9964 at exactly 792 bits on the COST2100 dataset using the described codebook-driven diffusion process would confirm or refute the central performance claim.","supporting_citations":[],"review_version":1}