- Rating
- 1
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 1
Read review
Paper summary
This paper investigates the correspondence between low-quality (LQ) and high-quality (HQ) images in a binary latent space. Built upon the Infinity backbone, the proposed method predicts bit-flip probabilities relative to LQ bits and incorporates quantization-margin conditioning, shared-prefix modeling, curriculum learning, and a scale-adaptive asymmetric focal loss. The authors frame their approach as preserving reliable content while correcting degradation-induced discrepancies, and evaluate it on four image super-resolution benchmarks.
Critical strengths
-
The analysis of LQ/HQ bit consistency across different scales, degradation levels, and quantization margins suggests that the quantization margin could serve as a useful conditioning signal. This is one of the most informative aspects of the paper.
-
GPABC achieves the highest MANIQA and CLIPIQA scores across all four datasets and the best LPIPS score on DIV2K-Val. Therefore, it would be inaccurate to dismiss the method as entirely ineffective. However, the key question is whether these improvements genuinely substantiate the authors' claims regarding selective bit correction and fidelity preservation.
Critical weaknesses
1.The proposed bitwise correction framework appears to be largely a reformulation of conventional bit prediction rather than a fundamentally different learning paradigm. While the additional conditioning mechanisms and training strategies may provide practical benefits, their individual contributions are not sufficiently isolated. Consequently, it remains unclear whether the reported improvements genuinely arise from the proposed correction-based formulation or primarily from auxiliary design choices. This significantly weakens the paper's central novelty claim.
2.Although the method achieves strong results on several no-reference perceptual quality metrics, its performance on full-reference metrics, particularly on real-world benchmarks, is substantially worse than that of competing methods. This raises serious concerns about whether the proposed approach actually preserves reliable image content, as claimed, or simply favors perceptually appealing but potentially less faithful reconstructions. Moreover, the paper lacks direct evidence demonstrating that its correction mechanism effectively identifies and preserves reliable bits. As a result, the experimental findings do not convincingly support the claimed advantages in balancing perceptual quality and reconstruction fidelity.
Questions for authors
1.What is the fundamental methodological contribution of the proposed correction-based formulation? Given that the basic bit-flip prediction objective can be reformulated as conventional HQ-bit prediction under identical conditions, it is unclear whether the correction formulation itself introduces a meaningful algorithmic advantage. Could the authors clarify what fundamentally distinguishes GPABC from standard conditional bit prediction and provide controlled comparisons that isolate the contributions of the correction formulation, shared-prefix modeling, margin conditioning, and the proposed loss?
2.Does the proposed framework actually perform selective correction as claimed? The current analysis primarily characterizes statistical relationships between LQ and HQ representations, but does not directly establish that the trained model preserves reliable bits while selectively correcting erroneous ones. Could the authors provide quantitative evidence of actual correction behavior during autoregressive inference, including false-flip and missed-flip rates across scales, and demonstrate improvements over a matched direct bit-prediction baseline? Without such evidence, the central correction-based interpretation remains insufficiently validated.
3.How can the claimed fidelity-preservation and perceptual-quality advantages be reconciled with the substantially worse full-reference results on RealSR and DRealSR? GPABC underperforms competing methods across PSNR, SSIM, LPIPS, and DISTS on these benchmarks, despite achieving stronger no-reference quality scores. Could the authors demonstrate that GPABC achieves a genuinely better perception–distortion trade-off under matched evaluation conditions? Otherwise, the current results appear insufficient to support the claimed advantages in preserving reliable image content.
8 comments
Author Response – Part 2
User Study of Reconstruction Fidelity. We conducted a user study with 15 participants using 50 image groups randomly sampled from DIV2K-Val, RealSR, and DRealSR. Each group includes LQ, HQ, and the corresponding outputs from AR-based methods, DM-based methods, and our GPABC. All outputs were presented anonymously in randomized order. For each image group, all participants were asked to select the output with (i) the highest fidelity to HQ, (ii) the best visual quality, and (iii) the best overall balance between fidelity and perceptual quality, yielding 750 votes per criterion. As shown in Table R2, GPABC leads across all three evaluation dimensions with vote shares of 45.20%, 41.07%, and 47.60%, which demonstrates that the additional detail generated by GPABC is preferable and remains faithful to the input.
Table R2. Human comparison results.
| Criterion | StableSR | OSEDiff | PiSA-SR | VARSR | PURE | DVAR | NSARM | GPABC |
|---|---|---|---|---|---|---|---|---|
| Fidelity | 3.60% | 6.00% | 11.07% | 5.33% | 6.93% | 6.13% | 15.73% | 45.20% |
| Quality | 3.20% | 3.07% | 6.67% | 4.93% | 12.80% | 12.27% | 16.00% | 41.07% |
| Overall | 3.20% | 4.53% | 9.33% | 4.80% | 7.33% | 6.00% | 17.20% | 47.60% |
Analysis of FR-IQA Scores and Fidelity–Perception Balance. To understand the lower FR-IQA scores on RealSR and DRealSR, we evaluated the HQ reference images themselves using NR-IQA metrics. As shown in Table R3, the HQ reference images exhibit relatively low NR-IQA scores, particularly on RealSR and DRealSR, as they often retain mild optical blur or sensor noise. Consequently, when a generative restoration model removes these imperfections and synthesizes sharper details, its similarity to the HQ decreases. This partially explains the lower FR-IQA scores of GPABC on RealSR and DRealSR. We think it should be taken into consideration that HQ references are not necessarily ideal targets, as their imperfections may also contribute to the discrepancy between FR-IQA and NR-IQA metrics.
Table R3. NR-IQA scores of the HQ reference images.
| Dataset | MANIQA ↑ | CLIPIQA ↑ | MUSIQ ↑ | TOPIQ ↑ |
|---|---|---|---|---|
| DIV2K-Val | 0.3884 | 0.6204 | 65.55 | 0.5776 |
| RealSR | 0.3295 | 0.4590 | 57.46 | 0.4140 |
| DRealSR | 0.3152 | 0.4751 | 50.56 | 0.3932 |
To further verify the perception–distortion trade-off under matched evaluation conditions, we trained a GPABC variant, GPABC-M, using milder synthetic degradation parameters so that the sharpness of its restored images aligns more closely with the HQ references. As shown in the table, GPABC-M reaches a PSNR of 25.45, which is comparable to the 25.50 of PiSA-SR, while achieving a higher MANIQA score of 0.5753 compared with 0.4810 of PiSA-SR. Moreover, compared with AR-based methods such as PURE and DVAR at comparable NR-IQA levels, GPABC-M achieves higher reconstruction fidelity across FR-IQA metrics, improving PSNR to 25.45 compared with 23.38 of PURE and 23.57 of DVAR. These comparisons support that our method achieves a favorable fidelity–perception balance.
Table R4. Comparison of all methods and GPABC-M on real-world RealSR benchmark.
| Metric | StableSR | SeeSR | SUPIR | OSEDiff | PiSA-SR | VARSR | VARestorer | PURE | DVAR | NSARM | GPABC | GPABC-M |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PSNR ↑ | 24.80 | 25.15 | 22.97 | 25.15 | 25.50 | 25.59 | 24.91 | 23.38 | 23.57 | 23.54 | 23.54 | 25.45 |
| SSIM ↑ | 0.7186 | 0.7210 | 0.5515 | 0.7341 | 0.7418 | 0.7266 | 0.6906 | 0.6502 | 0.6751 | 0.6412 | 0.6396 | 0.7357 |
| LPIPS ↓ | 0.2973 | 0.3008 | 0.4816 | 0.2920 | 0.2672 | 0.3225 | 0.2733 | 0.3496 | 0.3220 | 0.3694 | 0.4069 | 0.3056 |
| DISTS ↓ | 0.2159 | 0.2224 | 0.2717 | 0.2128 | 0.2044 | 0.2358 | 0.2188 | 0.2348 | 0.2315 | 0.2547 | 0.2688 | 0.2249 |
| MANIQA ↑ | 0.4422 | 0.5437 | 0.4335 | 0.4717 | 0.4810 | 0.5482 | 0.4437 | 0.4495 | 0.5556 | 0.6046 | 0.6695 | 0.5753 |
| CLIPIQA ↑ | 0.6199 | 0.6703 | 0.5742 | 0.6688 | 0.6704 | 0.6990 | 0.6148 | 0.6623 | 0.6823 | 0.7356 | 0.7831 | 0.7217 |
| MUSIQ ↑ | 66.09 | 69.82 | 61.00 | 69.09 | 70.15 | 71.23 | 68.54 | 65.42 | 70.30 | 71.13 | 70.66 | 69.29 |
| TOPIQ ↑ | 0.5899 | 0.6888 | 0.5651 | 0.6248 | 0.6369 | 0.6990 | 0.6175 | 0.5926 | 0.6982 | 0.7315 | 0.7266 | 0.7001 |
Identification and Preservation of Reliable Bits. To directly verify that GPABC identifies and preserves reliable bits, Table R5 reports the preservation rate (the proportion of consistent LQ–HQ bits kept unchanged) across all 13 scales on RealSR. As can be seen, at coarse scales (s1–s4) encoding global structures and low-frequency content, GPABC preserves over 85% of reliable bits (up to 88.69% at s1 and 88.68% at s2), while moderately decreasing to 67.98%–79.58% at finer scales (s8–s13) as our scale-adaptive loss actively encourages bit flips to synthesize missing high-frequency details.
Table R5. Per-scale preservation rates on RealSR.
| Metric | s1 | s2 | s3 | s4 | s5 | s6 | s7 | s8 | s9 | s10 | s11 | s12 | s13 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Preservation rate ↑ | 0.8869 | 0.8868 | 0.8720 | 0.8509 | 0.8416 | 0.8337 | 0.8282 | 0.7958 | 0.7710 | 0.7557 | 0.7151 | 0.6798 | 0.6975 |
Author Response – Part 3
Q2. Selective Correction Behavior During Autoregressive Inference
We thank the reviewer for this insightful comment. Following your suggestion, we evaluated the actual bitwise correction behavior of GPABC during autoregressive inference on the real-world RealSR benchmark and compared it against the matched prediction-style baseline. At each scale , we measure three metrics. Bit agreement rate measures the proportion of predicted bits that match the target HQ bits. False-flip rate measures the proportion of bits that the model erroneously flips among reliable LQ bits that agree with HQ. Missed-flip rate measures the proportion of bits that the model fails to flip among erroneous LQ bits that disagree with HQ. Table R6 reports these metrics across all scales, revealing three findings that directly validate our selective correction claim:
-
GPABC achieves higher bit agreement rates across all scales: By explicitly conditioning on current-scale LQ bits and quantization margins as correction anchors, GPABC accurately aligns degraded representations with their HQ targets and achieves higher bit agreement rates across all 13 scales from s1 to s13, for instance improving s1 from 0.8387 to 0.8784, s2 from 0.7891 to 0.8647, s8 from 0.6617 to 0.7342, and s13 from 0.5978 to 0.6315.
-
GPABC preserves reliable bits with lower false-flip rates across all scales: Across all scales, especially coarse scales that encode global structures and low-frequency content, GPABC achieves lower false-flip rates than the prediction-style baseline. For example, GPABC reduces the false-flip rate from 0.1623 to 0.1131 at s1, from 0.2055 to 0.1132 at s2, and from 0.3741 to 0.3025 at s13. This demonstrates that GPABC effectively identifies and preserves reliable LQ bits, retaining over 88.6% of reliable bits at s1 and s2 and over 82.8% across scales s1 to s7.
-
GPABC exhibits a relatively higher missed-flip rate across all scales: Without current-scale LQ anchors, the prediction-style baseline flips bits more aggressively, resulting in a lower missed-flip rate, such as 0.3743 vs. 0.4150 at s7, but at the cost of a substantially higher false-flip rate, such as 0.2607 vs. 0.1718, consequently leading to a lower bit agreement rate of 0.7103 vs. 0.7691. In contrast, GPABC performs selective correction, which ultimately yields superior overall bit agreement rates across every scale.
Table R6. Per-scale correction behavior during autoregressive inference on RealSR.
| Metric | Method | s1 | s2 | s3 | s4 | s5 | s6 | s7 | s8 | s9 | s10 | s11 | s12 | s13 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Bit agreement rate ↑ | Baseline | 0.8387 | 0.7891 | 0.7751 | 0.7388 | 0.7234 | 0.7116 | 0.7103 | 0.6617 | 0.6457 | 0.6423 | 0.6019 | 0.5927 | 0.5978 |
| Bit agreement rate ↑ | GPABC | 0.8784 | 0.8647 | 0.8424 | 0.8104 | 0.7970 | 0.7824 | 0.7691 | 0.7342 | 0.7095 | 0.6868 | 0.6536 | 0.6259 | 0.6315 |
| False-flip rate ↓ | Baseline | 0.1623 | 0.2055 | 0.2129 | 0.2477 | 0.2605 | 0.2692 | 0.2607 | 0.3181 | 0.3327 | 0.3260 | 0.3771 | 0.3865 | 0.3741 |
| False-flip rate ↓ | GPABC | 0.1131 | 0.1132 | 0.1280 | 0.1491 | 0.1584 | 0.1663 | 0.1718 | 0.2042 | 0.2290 | 0.2443 | 0.2849 | 0.3202 | 0.3025 |
| Missed-flip rate ↓ | Baseline | 0.1443 | 0.2323 | 0.2682 | 0.3042 | 0.3255 | 0.3449 | 0.3743 | 0.3907 | 0.4087 | 0.4347 | 0.4472 | 0.4551 | 0.4635 |
| Missed-flip rate ↓ | GPABC | 0.1593 | 0.2348 | 0.2793 | 0.3396 | 0.3594 | 0.3861 | 0.4150 | 0.4363 | 0.4409 | 0.4715 | 0.4731 | 0.4760 | 0.4892 |
We will add this per-scale correction behavior analysis during autoregressive inference on RealSR to the appendix of the revised manuscript.
Author Response – Part 1
We sincerely thank the reviewer for recognizing our informative bit-level analysis, and for providing insightful comments including the constructive feedback allowing us to improve our work.
W1 & Q1. Methodological Contributions and Component Ablations
We thank the reviewer for this insightful comment. Under identical conditioning and training configurations, the LQ-relative flip prediction objective and direct HQ-bit prediction are mathematically equivalent. Specifically, for a single bit, let b be the binary LQ bit where sign(qLQ)=2b−1, y the HQ target, and z=l1−l0 the logit difference between the predicted positive (l1) and negative (l0) bit values. Following Equation (7) in the paper, the flip logit is formulated as sign(qLQ)(l0−l1)=(1−2b)z, and the target y is mapped to the flip label b⊕y. This mapping preserves the BCE loss and its gradients.
Although mathematically equivalent, the correction perspective better characterizes the nature of image restoration, whose core goal, unlike image generation, is to preserve reliable content already present in the LQ input while correcting degradation-induced errors. This correction perspective directly inspired us to design LQ-anchored bit and margin conditioning. GPABC injects embeddings of the current-scale LQ bits and their quantization margins to preserve bits consistent with HQ while flipping inconsistent ones. In contrast, under the prediction perspective, existing autoregressive SR methods (e.g., VARSR, PURE, and DVAR) condition on the LQ image via LQ-derived [SOS] and context tokens, without taking the current-scale LQ information as explicit inputs. Table R1 compares this existing LQ conditioning with our LQ-anchored bit and margin conditioning on RealSR, with all other components and settings held fixed. As can be seen, our LQ-anchored conditioning improves both pixel-level fidelity and perceptual quality, e.g., increasing PSNR from 23.07 to 23.54 and MANIQA from 0.6389 to 0.6695. These results verify the advantage of explicit LQ-anchored bit and margin conditioning over existing LQ conditioning strategies.
To further demonstrate the effectiveness of each component, we conduct controlled ablation experiments isolating individual components. Table R1 shows that removing the quantization-margin embedding (w/o margin embedding) deprives the model of bit-reliability cues. The model may over-flip reliable bits under LSAF, which slightly increases certain NR-IQA scores but leads to a degradation across all full-reference fidelity metrics. Removing LSAF (w/o LSAF) causes the model to tend toward conservative bit preservation, as unchanged bits dominate across scales and are easier to learn than bit flips. While this slightly improves reference-based metrics, it leads to a drop across all perceptual quality metrics. In addition, using independent encoding (w/o shared prefix) introduces accumulated context shifts across scales and leads to consistent degradation across all metrics, with PSNR dropping to 23.28 and CLIPIQA dropping to 0.7048.
Table R1. Ablations of GPABC’s conditioning and design components on RealSR. The variant without margin embedding retains the current-scale LQ-bit input. The best result in each column is bolded.
W2 & Q3. Content Preservation and Fidelity–Perception Balance
We appreciate the reviewer’s insightful comments. To address your concern that GPABC may favor perceptually appealing but less faithful reconstructions, we conducted a user study that evaluates fidelity to HQ as a separate criterion. We further examined the perception–distortion trade-off of GPABC under matched evaluation conditions, and we also provide direct evidence on the preservation of reliable bits.