mollangdata/LangMolDiode-Qwen3.5-4B-SFT
Image-Text-to-Text • 5B • Updated • 21
Datasets:
id stringlengths 10 10 | molecule_id stringlengths 3 9 | smiles stringlengths 4 524 | system_prompt stringclasses 1
value | input_prompt stringlengths 837 5.64k | generated_response stringlengths 23 76.8k | generated_reasoning stringlengths 180 237k | generated_smiles stringlengths 4 517 | sft_response stringlengths 287 156k | generation_model stringclasses 2
values | reasoning_effort stringclasses 1
value | matches_smiles bool 1
class |
|---|---|---|---|---|---|---|---|---|---|---|---|
sft_000001 | 70818717 | CCS(=O)(=O)NC1CCOCC1COc1ccc(-c2cccnc2)cc1 | You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous molecular structure description, return the corresponding SMILES string enclosed in <smiles> and </smiles> tags. | Analyze the given molecule structure description and generate the corresponding molecular structure as SMILES representation.
- This description provides an unambiguous and detailed account sufficient to completely reconstruct the molecule.
- Interpret stereochemistry, charges, and isotopes if mentioned in the descrip... | Based on the detailed structural description provided, here is the step-by-step construction of the molecule:
1. **Ethanesulfonamide unit**: An ethyl group bonded to a sulfonyl group, which is bonded to an amino group. This gives the fragment `CCS(=O)(=O)N`.
2. **Tetrahydropyran (oxane) ring**: A saturated six-membere... | The goal is to generate a SMILES string based on the detailed molecule description.
Let's break down the description step-by-step:
1. **Ethanesulfonamide unit:**
* Ethyl group bonded to sulfonyl sulfur: `CCS(=O)(=O)N`
* Specifically: `C-C-S(=O)(=O)-N-`
2. **Sulfonamide nitrogen attachment:**
* Bo... | O1CC(COc2ccc(cc2)c3cnccc3)C(NS(=O)(=O)CC)CC1 | <think>
The goal is to generate a SMILES string based on the detailed molecule description.
Let's break down the description step-by-step:
1. **Ethanesulfonamide unit:**
* Ethyl group bonded to sulfonyl sulfur: `CCS(=O)(=O)N`
* Specifically: `C-C-S(=O)(=O)-N-`
2. **Sulfonamide nitrogen attachment:**
... | glm-5.1-fp8 | high | true |
sft_000002 | 7714012 | COc1cc(OC)cc(OCc2nc(-c3ccccc3C)no2)c1 | You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous molecular structure description, return the corresponding SMILES string enclosed in <smiles> and </smiles> tags. | Analyze the given molecule structure description and generate the corresponding molecular structure as SMILES representation.
- This description provides an unambiguous and detailed account sufficient to completely reconstruct the molecule.
- Interpret stereochemistry, charges, and isotopes if mentioned in the descrip... | <thinking>
The request asks for the SMILES representation of a molecule based on a detailed structural description.
1. **Central Core**: 1,2,4-oxadiazole ring, aromatic.
* Atom order specified: O1–N2–C3–N4–C5.
* SMILES for this core: `o1ncnc1` (where `o1` is O1, `n` is N2, `c` is C3, `n` is N4, and `c1` i... | The user wants a SMILES string based on a detailed molecule description.
Central core: 1,2,4-oxadiazole ring. Atoms in order: O1-N2-C3-N4-C5.
C3 substituent: 2-methylphenyl (o-tolyl) group. C3 is bonded to B1 (ipso). B2 (ortho to B1) has a methyl group.
C5 substituent: methylene linker (-CH2-). CH2 is bonded to an ethe... | o1nc(-c2c(C)cccc2)nc1COc3cc(OC)cc(OC)c3 | <think>
The user wants a SMILES string based on a detailed molecule description.
Central core: 1,2,4-oxadiazole ring. Atoms in order: O1-N2-C3-N4-C5.
C3 substituent: 2-methylphenyl (o-tolyl) group. C3 is bonded to B1 (ipso). B2 (ortho to B1) has a methyl group.
C5 substituent: methylene linker (-CH2-). CH2 is bonded to... | glm-5.1-fp8 | high | true |
sft_000003 | 63298954 | CC(C)CNC(=O)C(C)NCC(O)C1CC1 | You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous molecular structure description, return the corresponding SMILES string enclosed in <smiles> and </smiles> tags. | Analyze the given molecule structure description and generate the corresponding molecular structure as SMILES representation.
- This description provides an unambiguous and detailed account sufficient to completely reconstruct the molecule.
- Interpret stereochemistry, charges, and isotopes if mentioned in the descrip... | Based on the detailed description, we can construct the molecule step by step:
1. **Propanamide framework**: The core is a 2-aminopropanamide. This starts with an amide carbonyl `C(=O)`. Attached to it is the α-carbon, which is an sp³ carbon bearing a methyl group and an amino group. This gives us `C(=O)C(C)N`.
2. *... | The user wants to generate the SMILES representation of a molecule based on a detailed description.
Let's break down the description step by step:
1. **Core framework:** Three-carbon propanamide framework.
* Start with amide carbonyl carbon: `C(=O)`
* Attach amide nitrogen on one side: `NC(=O)` -> wait, ... | O=C(C(C)NCC(O)C1CC1)NCC(C)C | <think>
The user wants to generate the SMILES representation of a molecule based on a detailed description.
Let's break down the description step by step:
1. **Core framework:** Three-carbon propanamide framework.
* Start with amide carbonyl carbon: `C(=O)`
* Attach amide nitrogen on one side: `NC(=O)` -... | glm-5.1-fp8 | high | true |
sft_000004 | 116512936 | CCCCN=C(NN)N(CC)c1ccccc1C | You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous molecular structure description, return the corresponding SMILES string enclosed in <smiles> and </smiles> tags. | Analyze the given molecule structure description and generate the corresponding molecular structure as SMILES representation.
- This description provides an unambiguous and detailed account sufficient to completely reconstruct the molecule.
- Interpret stereochemistry, charges, and isotopes if mentioned in the descrip... | <smiles>CCCCN=C(N(CC)c1c(C)cccc1)NN</smiles> | The user wants a SMILES representation of a molecule based on a detailed description.
Let's break down the description:
1. Central carbon (guanidine carbon) C0 bonded to three nitrogens N1, N2, N3.
2. C0 is double bonded to N1.
3. C0 is single bonded to N2 and N3.
4. N1 substituent: n-butyl group (CCCC). Since N1 has a... | CCCCN=C(N(CC)c1c(C)cccc1)NN | <think>
The user wants a SMILES representation of a molecule based on a detailed description.
Let's break down the description:
1. Central carbon (guanidine carbon) C0 bonded to three nitrogens N1, N2, N3.
2. C0 is double bonded to N1.
3. C0 is single bonded to N2 and N3.
4. N1 substituent: n-butyl group (CCCC). Since ... | glm-5.1-fp8 | high | true |
sft_000005 | 61970535 | CCNC(C)(CCN(C)c1ccc(F)cc1)C(=O)OC | You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous molecular structure description, return the corresponding SMILES string enclosed in <smiles> and </smiles> tags. | Analyze the given molecule structure description and generate the corresponding molecular structure as SMILES representation.
- This description provides an unambiguous and detailed account sufficient to completely reconstruct the molecule.
- Interpret stereochemistry, charges, and isotopes if mentioned in the descrip... | Based on the detailed description provided, we can systematically construct the molecule and derive its SMILES representation:
1. **Methyl Ester and Carbonyl Carbon (C1):** The description specifies a methyl ester at the terminal carboxyl group. The carbonyl carbon is C1, double-bonded to an oxygen and single-bonded ... | The user wants to generate a SMILES string based on a detailed description of a molecule.
1. **Identify the core structure:** A four-carbon carboxylic-acid derivative (butanoate) with a methyl ester at one end.
2. **Ester group:** Carbonyl carbon (C1) double bonded to O, single bonded to -O-CH3. So, C1 is part of -C... | COC(=O)C(C)(NCC)CCN(C)c1ccc(F)cc1 | <think>
The user wants to generate a SMILES string based on a detailed description of a molecule.
1. **Identify the core structure:** A four-carbon carboxylic-acid derivative (butanoate) with a methyl ester at one end.
2. **Ester group:** Carbonyl carbon (C1) double bonded to O, single bonded to -O-CH3. So, C1 is pa... | glm-5.1-fp8 | high | true |
sft_000006 | 120513683 | CCOc1ccccc1C(F)(F)C(=O)N(C)CC(=O)O | "You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous m(...TRUNCATED) | "Analyze the given molecule structure description and generate the corresponding molecular structure(...TRUNCATED) | <smiles>OC(=O)CN(C)C(=O)C(F)(F)c1c(OCC)cccc1</smiles> | "The user wants a SMILES representation based on a detailed molecule description.\n\nLet's break dow(...TRUNCATED) | OC(=O)CN(C)C(=O)C(F)(F)c1c(OCC)cccc1 | "<think>\nThe user wants a SMILES representation based on a detailed molecule description.\n\nLet's (...TRUNCATED) | glm-5.1-fp8 | high | true |
sft_000007 | 119683177 | CCCC(C)(N)C(=O)N(C)CC(=O)Nc1ccc(Cl)c(C(F)(F)F)c1 | "You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous m(...TRUNCATED) | "Analyze the given molecule structure description and generate the corresponding molecular structure(...TRUNCATED) | "Based on the detailed description, we can systematically construct the SMILES string for the molecu(...TRUNCATED) | "The user wants to convert a detailed molecule structure description into a SMILES string.\n\nLet's (...TRUNCATED) | CCCC(C)(N)C(=O)N(C)CC(=O)Nc1cc(C(F)(F)F)c(Cl)cc1 | "<think>\nThe user wants to convert a detailed molecule structure description into a SMILES string.\(...TRUNCATED) | glm-5.1-fp8 | high | true |
sft_000008 | 22294160 | CN(CC1CCC(OCCCCO)CC1)S(=O)(=O)c1ccc(Cl)cc1 | "You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous m(...TRUNCATED) | "Analyze the given molecule structure description and generate the corresponding molecular structure(...TRUNCATED) | "Based on the detailed description, we can construct the molecule step-by-step:\n\n1. **Benzene ring(...TRUNCATED) | "The user wants a SMILES representation of a molecule based on a detailed description.\nLet's break (...TRUNCATED) | Clc1ccc(cc1)S(=O)(=O)N(C)CC2CCC(OCCCCO)CC2 | "<think>\nThe user wants a SMILES representation of a molecule based on a detailed description.\nLet(...TRUNCATED) | glm-5.1-fp8 | high | true |
sft_000009 | 5720906 | Cc1nc(Cl)cc(N/N=C\c2ccc([N+](=O)[O-])cc2)n1 | "You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous m(...TRUNCATED) | "Analyze the given molecule structure description and generate the corresponding molecular structure(...TRUNCATED) | "The molecule is constructed based on a pyrimidine ring with substituents at specific positions, lea(...TRUNCATED) | "The molecule is based on a pyrimidine ring.\nRing numbering: N1-C2-N3-C4-C5-C6.\nC2 has a methyl gr(...TRUNCATED) | Cc1nc([NH]/N=C\c2ccc([N+]([O-])=O)cc2)cc(Cl)n1 | "<think>\nThe molecule is based on a pyrimidine ring.\nRing numbering: N1-C2-N3-C4-C5-C6.\nC2 has a (...TRUNCATED) | glm-5.1-fp8 | high | true |
sft_000010 | 173397138 | O=CN(C(=O)NC1CCCCC1)c1ccc(Cl)c(Cl)c1 | "You are a chemistry assistant specialized in molecular structure generation. Given an unambiguous m(...TRUNCATED) | "Analyze the given molecule structure description and generate the corresponding molecular structure(...TRUNCATED) | "Based on the detailed description, we can break down the molecule into its constituent parts and as(...TRUNCATED) | "The user wants a SMILES representation of a molecule based on a detailed description.\n\nLet's brea(...TRUNCATED) | O=C(N(c1cc(Cl)c(Cl)cc1)C(=O)NC1CCCCC1) | "<think>\nThe user wants a SMILES representation of a molecule based on a detailed description.\n\nL(...TRUNCATED) | glm-5.1-fp8 | high | true |
End of preview. Expand in Data Studio
LangMolDiode SFT Corpus
LangMolDiode is a language-conditional molecule generator that generates SMILES from detailed molecular structure descriptions.
Anonymous project code: LangMolDiode
This corpus contains 75,666 examples from the filtered Qwen3.5-4B SFT run (maximum training sequence length: 40,960 tokens). Each row is one training example, in original training order.
Columns
| Column | Description |
|---|---|
id |
Release-local row identifier. |
molecule_id |
Source molecule identifier when available. |
smiles |
Verified reference SMILES used as the final SFT answer. |
system_prompt |
Exact system message used for SFT. |
input_prompt |
Exact user message used for SFT. |
generated_response |
Raw visible response returned by the teacher model, corresponding to the API message content. |
generated_reasoning |
Separate reasoning field returned by the teacher model. |
generated_smiles |
Candidate SMILES selected from the tagged SMILES in generated_response during curation. This preserves the teacher's generated string and may differ textually from smiles. |
sft_response |
Exact assistant message used for SFT: normalized generated_reasoning inside <think> tags, followed by the verified reference smiles inside <smiles> tags. |
generation_model |
Teacher model that generated the response. |
reasoning_effort |
Effective reasoning-effort setting recorded for the generation. |
matches_smiles |
Whether generated_smiles and smiles are chemically equivalent under the stereochemistry-aware RDKit validation. |
- Downloads last month
- 50