- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
LAIR is an agentic image restoration framework that retrieves and reuses historical restoration trajectories using fine-grained degradation representations. Its DQ-MCE encoder combines learnable degradation queries, cross-content momentum contrastive learning, and synthetic degradation-parameter regression. Retrieved trajectories are filtered by degradation compatibility, ranked by stored quality, and verified after execution. The system falls back to AgenticIR planning when reuse fails and updates its memory with successful results. Experiments report modest fidelity improvements and substantial online runtime reductions on synthetic and real-world restoration data.
Critical strengths
-
The motivation is practical and well supported. Coarse degradation labels can group images that require different restoration tools. Learning finer representations for experience retrieval directly addresses this limitation.
-
LAIR reduces runtime from 156.91 to 65.90 seconds per image in the reported synthetic comparison. Its integration into 4KAgent and evaluation on real-world images provide additional evidence that memory reuse can reduce execution cost.
Critical weaknesses
-
The novelty over existing memory-based agents is incremental. SEAR already retrieves restoration trajectories using degradation-aware fingerprints. A matched comparison using the same memory, verification rules, and budget would better isolate the benefit of DQ-MCE.
-
The representation ablations are too narrow. Retrieval ablations cover only 100 rain-plus-dark images. Separate tests of degradation queries, cross-content pairing, momentum learning, and parameter-only training are needed to establish which components matter.
-
The online-memory evaluation protocol needs clarification. Updating memory during testing makes results potentially dependent on image order and previous queries. Shuffled-order repeats, memory-reset settings, and performance versus memory size would improve reproducibility.
-
The role of offline supervision and preparation cost is insufficiently isolated. Initial memories use reference-based selection of the best tool chains, while the encoder requires substantial synthetic training. Comparisons should match memory supervision and distinguish online savings from offline preparation costs.
Overall, I lean toward acceptance because the paper presents a coherent representation-learning approach with clear online efficiency benefits and useful supporting diagnostics. The contribution is incremental, but practically relevant. Stronger component ablations and a reproducible online-memory protocol would substantially strengthen the evidence.
Questions for authors
See weaknesses.
5 comments
W2. more ablation on DQ-MCE
We agree that the original 100-image rain+dark experiment provided limited coverage. We now report separate component ablations covering all four compound-degradation settings in Group B (400 images). For each split, 80 images per setting form the memory bank (320 total), and the remaining 20 per setting serve as restoration queries (80 total). We repeat evaluation over ten memory/query splits using seeds 0–9, keeping each variant’s trained checkpoint fixed and disabling planning fallback to assess memory reuse. The new tables report averages across these splits.
We evaluate five ablations: (i) query-based aggregation: replacing learnable queries and cross-attention with mean-pooled patch features and seven independent projection MLPs, approximately matching parameter count while preserving the output; (ii–iii) training objectives: using InfoNCE only or parameter-regression MSE only; (iv) cross-content pairing: selecting positives from another degraded version of the same clean crop, while retaining ordered-degradation matching and nearest-parameter pairing; and (v) EMA updates: setting while retaining stop-gradient keys and the negative queue; We also include Severity-L2, mean-patch, and CLS-token feature baselines.
| Extractor | PSNR ↑ | SSIM ↑ | Recall@5 (%) ↑ | Rank ↓ |
|---|---|---|---|---|
| DQ-MCE | ||||
| w/o queries | ||||
| Parameter only | ||||
| Contrastive only | ||||
| w/o cross-content | ||||
| w/o EMA | ||||
| Severity-L2 | ||||
| Mean patch | ||||
| CLS token |
The table shows the training objectives are necessary to help DQ-MCE extract fine-grained features, achiving better performance than discrete-labels.
W3. online-memory evaluation protocol
Our initial memory is constructed from 160 Group A exploration samples following the split of AgenticIR. These samples are excluded from the 1,440-image test set. Before each independent evaluation run, we restore exactly the same initial memory; online updates are retained within that run but are not carried over to subsequent runs.
The original results use a fixed ascending image order(from id 1 to id 100). To assess order sensitivity, we additionally evaluated all 1,440 test images under ten shuffled orders (seeds 0–9), changing only the test-image order.
| seed | PSNR-Y ↑ | SSIM-Y ↑ | LPIPS ↓ | MANIQA ↑ | CLIP-IQA ↑ | MUSIQ ↑ |
|---|---|---|---|---|---|---|
| 0 | 21.9693 | 0.6897 | 0.3332 | 0.3240 | 0.4815 | 58.0830 |
| 1 | 21.9465 | 0.6881 | 0.3336 | 0.3258 | 0.4837 | 58.3614 |
| 2 | 21.9354 | 0.6886 | 0.3341 | 0.3260 | 0.4812 | 58.5655 |
| 3 | 22.0371 | 0.6894 | 0.3350 | 0.3258 | 0.4813 | 58.1931 |
| 4 | 21.9241 | 0.6879 | 0.3345 | 0.3234 | 0.4813 | 58.2458 |
| 5 | 21.9139 | 0.6874 | 0.3350 | 0.3251 | 0.4819 | 58.3294 |
| 6 | 21.8586 | 0.6871 | 0.3377 | 0.3226 | 0.4811 | 58.0811 |
| 7 | 22.0419 | 0.6910 | 0.3313 | 0.3261 | 0.4808 | 58.3314 |
| 8 | 21.9828 | 0.6885 | 0.3347 | 0.3254 | 0.4817 | 58.1955 |
| 9 | 21.9957 | 0.6892 | 0.3345 | 0.3263 | 0.4814 | 58.2465 |
| mean ±std | 21.9605 ± 0.0568 | 0.6887 ± 0.0012 | 0.3344 ± 0.0016 | 0.3251 ± 0.0013 | 0.4816 ± 0.0008 | 58.2633 ± 0.1437 |
| Δ(main experiments − mean) | +0.0611 | −0.0002 | −0.0035 | +0.0015 | −0.0006 | +0.2446 |
These results show small aggregate metric variation across the tested permutations, which demonstrates our LAIR's robustness in shuffled-order repeats.
Regarding performance versus memory size, we have added a comparison of initial memory sizes (0, 80, and 160 entries) to the supplementary material (page 18 A.4), with online updates enabled in all settings. Memory initialization generally improves early-stage PSNR-Y compared with starting from an empty memory. The mean PSNR-Y gap between the 160- and 0-entry initializations decreases from 0.624 dB over images 1–40 to 0.240 dB over images 41–80. These results suggest reduced sensitivity to initialization during later processing because of the online memory growth, although the gaps fluctuate and some differences remain. The table below reports PSNR-Y (dB), averaged over successive, non-overlapping blocks of 10 images per degradation combination and then equally across the eight combinations.
| step | 0 | 80 | 160 |
|---|---|---|---|
| 1–10 | 22.5899 | 23.2701 | 23.2505 |
| 11–20 | 21.4530 | 21.5303 | 21.6369 |
| 21–30 | 21.2738 | 22.2120 | 22.3338 |
| 31–40 | 21.1354 | 21.0950 | 21.7263 |
| 41–50 | 23.7875 | 23.8368 | 24.0892 |
| 51–60 | 22.7229 | 22.7873 | 22.8520 |
| 61–70 | 22.4858 | 22.6889 | 22.5473 |
| 71–80 | 20.6479 | 21.1603 | 21.1156 |
W4. Clarification of preparation cost
LAIR and AgenticIR share the offline cost of exhaustive tool-chain execution on the same 160 exploration images to get inital memory(or static experience).The reported runtime reduction concerns online inference latency; it excludes offline preparation and encoder training.
However, for training cost, DQ-MCE requires an additional 2–3 hours of training on eight RTX A6000 GPUs, performed only once. But users benefit from roughly halved average waiting times and a more responsive restoration experience.
Thank you for recognizing LAIR’s efficiency gains and the evidence from the retrieval ablations.
W1. The comparison with SEAR
For this response, we independently implemented SEAR because we could not locate an official public implementation. We evaluated the independent implementation at the reported P-MCTS budget multiplier (=9[official parameter in paper]) and at (=7). The Group B comparison is shown below because it is also the split used for SEAR's published efficiency analysis, while
Severity-L2represents the variant that replaces only LAIR’s retrieval representation and matching with SEAR’s fingerprint-based retrieval, while keeping the initial memory, restoration tools, verification rules, planning budget, and memory-update policy unchanged.The tool call per image in official SEAR paper in Group B is 8.15 instead of 9.34 or 9.74 in our implementation. However, I guess the run time will also longer than LAIR. LAIR shows best fidelity and better efficiency.