cs.LGAug 19, 2026
SaveTraining Chemical Plausibility-Aware Large Language Models for Single-Step Retrosynthesis
Organizations: Insilico Medicine AI Limited, Abu Dhabi, UAE · Insilico Medicine Canada Inc., Montreal, Quebec, Canada · Insilico Medicine Hong Kong Ltd., Hong Kong SAR, China
Abstract
Single-step retrosynthesis is a central component of computer-aided synthesis planning, yet its intrinsically one-to-many nature is poorly captured by single-answer evaluation and benchmarking protocols. To address this, we introduce Top-K prompting as a robust training and inference paradigm to better capture diverse, plausible reaction predictions. We compile CREED-CCV-2+USPTO-XL, an ultra-large-scale dataset of ~45.6 million verified reactions to train the C3LM (Chemistry Constraint-Consistent Language Model). By integrating fine-tuning with ChemCensor-based and novelty-oriented rewards, our model achieves state-of-the-art performance on the OOD URSA-expert-2026 benchmark. Further analysis of reaction uniqueness shows that LLMs and conventional models explore complementary reaction spaces, motivating ensemble-based retrosynthesis systems. Overall, our results establish Top-K, plausibility-aware training as a practical new direction for robust future LLM-based synthesis planning.
Figures & tables
Figure 1: Comparison of representative top-performing models from each model tier, normalized by metric-specific frontier scores.
Figure 2: Data-generation pipeline for the CREED-CCV-2+USPTO-XL training set.
Figure 3: Top-1 Top-K transition on URSA-expert-2026 (Av. PT-Top-10 CC). Hollow = Top-1 task, filled = Top-K task; line length is the gain (right column, sorted).
| Model | URSA-expert-2026 | USPTO-50K-test-mini | ||||||
| Max | Av. PT-Top-K CC | Max | Av. PT-Top-K CC | |||||
| @3 | @5 | @10 | @3 | @5 | @10 | |||
| Proprietary Foundation Models | ||||||||
| Grok-4.1 | 1.86 | 1.56 | 1.31 | 0.86 | 3.99 | 2.61 | 2.02 | 1.24 |
| Grok-4.3 | 1.74 | 1.41 | 1.13 | 0.66 | 3.99 | 2.58 | 1.95 | 1.15 |
| Gemini 3.1 Pro | 1.91 | 1.69 | 1.46 | 1.08 | 4.32 | 2.92 | 2.34 | 1.59 |
Table 1: Plausibility-based evaluation in the single-step retrosynthesis Top- mode. Max : per-target maximum ChemCensor score averaged over TMs. Av. PT-Top-K CC : per-TM average ChemCensor score over top-K unique predictions; ChemCensor v1.1.1, USPTO-full as the source of synthetic precedents. See the full table in Appendix F . Results for recently released models are available at https://dddbench.insilico.com/ .
Figure 4: Intersection between the reactions predicted for the URSA-expert-2026 benchmark set by each LLM and those predicted by MHNreact. Values on the right ( ) show the average per-target difference between the number of reactions unique to the LLM and those unique to MHNreact; positive values indicate more LLM-unique reactions. C3LM-2 : C3LM-LFM2-CREED-CCV+USPTO ; C3LM-3 : C3LM-LFM2-CREED-CCV-2+USPTO-XL ; C3LM-5 : C3LM-LFM2-RFT-CC-NR . See all results in Figure 5 .
Appendix figures & tables8 assets
Supplementary material from the paper’s appendix.
Appendix
| Model Name | Short Name | Training Data | Reward |
| C3LM , Supervised Fine-Tuning, Top-1 task | |||
| C3LM-LFM2-CREED-CCV+USPTO * | C3LM-1 | TD-1 | – |
| C3LM , Supervised Fine-Tuning, Top-K task | |||
| C3LM-LFM2-CREED-CCV+USPTO | C3LM-2 | TD-1 | – |
| C3LM-LFM2-CREED-CCV-2+USPTO-XL | C3LM-3 | TD-2 | – |
| C3LM , Reinforcement Learning Fine-Tuning, Top-K task | |||
Table 2: C3LM family inventory. TD-1 = CREED-CCV + USPTO; TD-2 = CREED-CCV-2 + USPTO-XL. ∗ Trained in prior work ( Zagribelnyy et al., 2026b ) ; its predictions are re-scored under ChemCensor (v1.1.1) for comparison, without retraining.
| Model | URSA-expert-2026 | USPTO-50K-test-mini | ||||||
| Max | Av. PT-Top-K CC | Max | Av. PT-Top-K CC | |||||
| @3 | @5 | @10 | @3 | @5 | @10 | |||
| Proprietary Foundation Models | ||||||||
| Grok-4.1 | 1.73 | 1.28 | 0.92 | 0.47 | 4.01 | 2.29 | 1.53 | 0.79 |
| Grok-4.3 | 1.47 | 0.86 | 0.56 | 0.28 | 3.06 | 1.34 | 0.83 | 0.41 |
| Gemini 3.1 Pro | 1.63 | 0.92 | 0.57 | 0.28 | 4.00 | 1.70 | 1.03 | 0.51 |
Table 4: Plausibility-based evaluation in the single-step retrosynthesis Top- mode. Max : per-target maximum ChemCensor score averaged over TMs. Av. PT-Top-K CC : per-TM average ChemCensor score over top-K unique predictions; ChemCensor v1.1.1.
| Model | Intersection | Unique for Model | Unique for MHNreact | |
| C3LM-LFM2-CREED-CCV-2+USPTO-XL | 5.0 | 6.8 | 6.4 | +0.4 |
| C3LM-LFM2-RFT-CC-NR | 5.0 | 6.7 | 6.4 | +0.3 |
| C3LM-LFM2-RFT-CC | 4.8 | 6.0 | 6.6 | -0.6 |
| C3LM-LFM2-CREED-CCV+USPTO | 3.6 | 4.0 | 7.9 | -3.9 |
| Gemini 3.1 Pro | 5.2 | 3.5 | 6.2 | -2.7 |
| GPT 5.5 | 5.2 | 3.2 | 6.2 | -3.0 |
Table 5: Intersection of predicted reactions for URSA-expert-2026 benchmark set with the reference conventional model MHNreact. is the per-target difference between the number of reactions unique to the model and those unique to MHNreact.
Figure 5: Intersection of reactions predicted by LLMs and the conventional SSRS model MHNreact. Values on the right show the difference between the number of reactions unique to the model and those unique to MHNreact (model MHNreact). C3LM-1 : C3LM-LFM2-CREED-CCV+USPTO * ; C3LM-2 : C3LM-LFM2-CREED-CCV+USPTO ; C3LM-3 : C3LM-LFM2-CREED-CCV-2+USPTO-XL ; C3LM-4 : C3LM-LFM2-RFT-CC ; C3LM-5 : C3LM-LFM2-RFT-CC-NR .
Figure 6: Intersection of reactions predicted for X404-1768-5005 by C3LM-LFM2-CREED-CCV-2+USPTO-XL and MHNreact.
| Model | URSA-expert-2026 | USPTO-50K-test-mini | ||||||
| Max | Av. PT-Top-K CC | Max | Av. PT-Top-K CC | |||||
| @3 | @5 | @10 | @3 | @5 | @10 | |||
| Proprietary Foundation Models | ||||||||
| Grok-4.1 | 2.17 | 1.81 | 1.54 | 1.06 | 4.13 | 2.84 | 2.27 | 1.46 |
| Grok-4.3 | 2.02 | 1.62 | 1.32 | 0.82 | 4.14 | 2.81 | 2.19 | 1.35 |
| Gemini 3.1 Pro | 2.20 | 1.93 | 1.68 | 1.29 | 4.42 | 3.12 | 2.56 | 1.83 |
Table 6: Plausibility-based evaluation in the single-step retrosynthesis Top- mode. Max : per-target maximum ChemCensor score averaged over TMs. Av. PT-Top-K CC : per-TM average ChemCensor score over top-K unique predictions; ChemCensor v1.1.1, ChemCensor-U2P2 reference database (USPTO-full Pistachio Q3 2023) .
| Model | URSA-expert-2026 | USPTO-50K-test-mini | ||||||
| Max | Av. PT-Top-K CC | Max | Av. PT-Top-K CC | |||||
| @3 | @5 | @10 | @3 | @5 | @10 | |||
| Proprietary Foundation Models | ||||||||
| Grok-4.1 | 1.90 | 1.59 | 1.33 | 0.88 | 4.04 | 2.71 | 2.13 | 1.33 |
| Grok-4.3 | 1.80 | 1.45 | 1.16 | 0.69 | 4.02 | 2.65 | 2.03 | 1.22 |
| Gemini 3.1 Pro | 1.96 | 1.72 | 1.49 | 1.11 | 4.35 | 2.97 | 2.42 | 1.66 |
Table 7: Plausibility-based evaluation in the single-step retrosynthesis Top- mode. Max : per-target maximum ChemCensor score averaged over TMs. Av. PT-Top-K CC : per-TM average ChemCensor score over top-K unique predictions; ChemCensor v0.5.2.
Figure 7: Distribution of the number of reactant sets for products from USPTO-50K-test-mini. For each product, all reference reactant sets were retrieved by matching it against USPTO-full.