- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes CAMS-Diff, a coupled adaptive mixed-space diffusion framework for molecular translation. The method introduces two main components: (1) coupling token-wise adaptive Gaussian noising with discrete representation masking under a shared learned schedule, and (2) trajectory-informed timestep selection for efficient reduced-step inference. The framework is evaluated across molecule captioning, text-guided molecule generation, forward reaction prediction, retrosynthesis, and zero-shot generalization. Experimental results demonstrate improvements in molecular similarity and semantic reconstruction over several baselines. Ablation studies further support the effectiveness of the proposed coupling mechanism and adaptive timestep allocation.
Critical strengths
- Effective integration of complementary mechanisms. The coupling of adaptive Gaussian noising and discrete masking is well motivated. The ablation results in Table 5 demonstrate consistent improvements over individual and uncoupled variants, supporting the effectiveness of the proposed design.
- Comprehensive evaluation and promising efficiency. The method is evaluated across multiple molecular translation tasks and zero-shot transfer. It achieves competitive improvements in molecular similarity and demonstrates substantial inference acceleration through trajectory-informed reduced-step decoding.
Critical weaknesses
- Reduced molecular validity. CAMS-Diff achieves lower SMILES validity than several baselines, despite improvements in molecular similarity.
- Model capacity differences. Some baseline comparisons use different parameter counts, potentially affecting the fairness of performance comparisons.
- Additional profiling overhead. Trajectory-informed timestep selection requires full-step validation profiling, and this computational cost is excluded from the reported inference speedups.
- Limited theoretical justification. The timestep allocation theory relies on a continuous approximation, while the practical finite-step scheduling strategy remains heuristic.
Questions for authors
Please see the weaknesses.
7 comments
Response to W2
W2. Model Capacity Differences
To address the fairness concern, we performed additional capacity-controlled experiments, reducing CAMS-Diff from 180M to 125M parameters for text-guided molecule generation and from 91M to 60M for molecule captioning.
(1) Text-guided molecule generation (ChEBI-20).
We reduced CAMS-Diff to 125M parameters, matching UTGDiff and TGM-DLM. We additionally evaluated both CAMS-Diff configurations with the deterministic SMILES repair described in W1.
| Metric ↑ | BiMol-Diff (180M) | CAMS-Diff (125M) | 125M + repair | CAMS-Diff (180M) | 180M + repair |
|---|---|---|---|---|---|
| MACCS | 0.883 | 0.906 | 0.889 | 0.916 | 0.899 |
| RDK | 0.785 | 0.803 | 0.808 | 0.811 | 0.807 |
| Morgan | 0.758 | 0.782 | 0.772 | 0.790 | 0.780 |
| Validity | 0.812 | 0.748 | 0.823 | 0.748 | 0.823 |
Despite reducing parameter count by approximately 31%, the 125M configuration retains strong molecular similarity, outperforming the reproduced 180M BiMol-Diff baseline across all three fingerprint metrics. Moreover, SMILES repair increases validity to 0.823 for both CAMS-Diff configurations, exceeding BiMol-Diff's 0.812, while preserving superior fingerprint similarity.
(2) Molecule captioning (M3-20M).
We further reduced CAMS-Diff from 91M to 60M parameters, closely matching the reproduced 63M-parameter BiMol-Diff baseline.
| Metric ↑ | BiMol-Diff (63M) | CAMS-Diff (60M) | CAMS-Diff (91M) |
|---|---|---|---|
| BLEU | 0.557 | 0.600 | 0.610 |
| chrF++ | 0.710 | 0.709 | 0.705 |
| METEOR | 0.632 | 0.690 | 0.740 |
| BERTScore-F1 | 0.838 | 0.849 | 0.861 |
| MAUVE | 0.913 | 0.918 | 0.927 |
Despite having fewer parameters than BiMol-Diff, the 60M CAMS-Diff configuration outperforms it on four of the five captioning metrics, while achieving nearly identical chrF++ performance.
These capacity-controlled results complement our existing same-architecture component ablations, which isolate the effects of adaptive noising and coupled masking without changing parameter count (see our response to Reviewer CBtW, W2/Q1).
Together, these experiments demonstrate that CAMS-Diff retains its principal improvements at comparable or smaller model capacities, supporting the methodological contributions independently of model size. We will incorporate these experiments and include capacity-controlled comparisons for the remaining tasks in the revised manuscript.
Response to W3
W3. Additional Profiling Overhead
We thank the reviewer for raising this concern. Trajectory profiling is a one-time, offline procedure performed after training; it does not contribute to test-time generation latency. We quantify both its cost and the quality it buys.
(1) Profiling reuses the existing validation pass. Measuring denoising progress requires a full-step reverse pass over the validation set. This is the same pass already performed after training to assess generation quality; we additionally log position-wise reconstruction errors at each retained timestep, requiring no additional denoiser evaluations. Errors are accumulated as described in Appendix C.1.
(2) Remaining stages require no denoiser evaluations. Let be the validation-set size, the full budget, the retained target positions, the reduced budget, the output length, the prediction dimension, and the cost of one denoiser evaluation.
| Stage | Operation | Time complexity |
|---|---|---|
| 1. Denoising progress | Full-step validation and error accumulation | , shared with validation |
| 2. Local importance | Fit error curves, compute local importance | |
| 3. Shared importance | Aggregate into | |
| 4. Timestep allocation | Select interior steps from bins |
Once validation statistics are available, the incremental scheduling cost is
(3) Measured overhead (, single NVIDIA A100).
| Task | Positions | Stage 2 | Stage 3 | Stage 4 | Total |
|---|---|---|---|---|---|
| Captioning | 64 | 104.1 ms | 0.6 ms | 0.1 ms | 0.105 s |
| Text-guided generation | 256 | 406.2 ms | 2.3 ms | 0.1 ms | 0.409 s |
| Forward prediction | 128 | 204.8 ms | 1.2 ms | 0.1 ms | 0.206 s |
| Retrosynthesis | 128 | 204.8 ms | 1.2 ms | 0.1 ms | 0.206 s |
(4) Overhead relative to generation cost, and the quality it buys. The scheduling cost is incurred once per task and does not scale with the number of samples generated. We report it against the cost of generating 100 samples, alongside the quality obtained at the same denoising budget.
| Task | Sampling / sample | / 100 samples | Overhead | Share at 100 | Share at 1,000 | Quality gain at |
|---|---|---|---|---|---|---|
| Captioning | 0.0091 s | 0.91 s | 0.105 s | 11.5% | 1.2% | +1.00 BLEU pp (DDIM +0.80) |
| Text-guided generation | 0.079 s | 7.90 s | 0.409 s | 5.2% | 0.5% | +0.67 MACCS pp (DDIM +0.40) |
| Forward prediction | 0.102 s | 10.20 s | 0.206 s | 2.0% | 0.2% | +0.50 validity pp (DDIM +1.50) |
| Retrosynthesis | 0.080 s | 8.00 s | 0.206 s | 2.6% | 0.3% | +1.90 validity pp (DDIM +1.80) |
Importantly, the two arms consume identical sampling compute: both perform exactly denoiser evaluations per sample and differ only in where those evaluations are placed along the reverse trajectory. The overhead is therefore not an inference-time latency trade-off, but a one-time offline cost after which every generated sample is improved at no recurring expense. It amortizes to below 1.5% within 1,000 samples on all four tasks, and the same importance profile is reused across sampling budgets , with the constructed schedule fixed across all test examples.
(5) On the reported speedups. These measure test-time generation latency. Since schedule construction is performed once offline and the retained timesteps are fixed thereafter, inference requires exactly denoiser evaluations per sample with no recurring profiling cost. We will state this distinction, the stage-wise complexity, and the measured overhead explicitly in Appendix C.1 of the revised manuscript.
Response to W1
W1. Reduced Molecular Validity
We thank the reviewer for highlighting this concern. We acknowledge the lower raw SMILES validity and have conducted additional error analysis and deterministic repair experiments, detailed in our response to Reviewer CBtW (W1).
As shown in Figure 10 (Appendix G), invalid generations are dominated by SMILES serialization errors, including unpaired ring labels (43.95%), unclosed branches (41.35%), and extra closing parentheses (37.86%), whereas only 4.87% involve invalid atomic valence (categories may overlap).
Importantly, a simple, deterministic correction of missing closing parentheses recovers 248 of 3,300 predictions, improving validity from 0.748 to 0.823 (+7.5 percentage points) without retraining. This exceeds the uncorrected diffusion baselines in Table 3, while retaining strong MACCS similarity (0.899) and unchanged exact match (0.232).
Furthermore, as discussed in Reviewer CBtW (W1), validity alone does not establish correct molecular generation. CAMS-Diff achieves higher exact match on retrosynthesis, forward prediction, and zero-shot generation (Tables 4 and 6), supporting its downstream effectiveness despite lower raw validity.
We refer the reviewer to our detailed response to CBtW (W1) for the matched-validity evaluation, error analysis, and repair results, which we will incorporate into the revised manuscript.