S2-TOMG-Bench

Nine chemistry generation subtasks (45,000 instructions) that ask for SMILES molecules satisfying atom, edit, or property constraints, scored with RDKit success and weighted success.

Also known as: S²-Bench, S2-Bench, TOMG-Bench, Speak-to-Structure

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorydomain
Subcategoryopen-domain natural-language molecule generation (MolCustom, MolEdit, MolOpt)
Page statusactive
Metricweighted success rate (WSR); success rate (SR) and validity also reported
Directionhigher_is_better
Unit%
Dataset size45000
Dataset licenceApache-2.0
PublisherAuthors of Speak-to-Structure (KDD 2026); Hugging Face dataset phenixace

What it measures

S2-TOMG-Bench (Speak-to-Structure; also called S²-Bench) asks a model to invent a molecule from an English instruction and return it as SMILES, usually as JSON `{"molecule": "..."}`. Three families probe different skills: MolCustom (build from atom, bond, or functional-group counts), MolEdit (add, delete, or substitute a group on a source molecule), and MolOpt (raise or lower LogP, molar refractivity, or QED). The point is one-to-many generation, not retrieving a single stored SMILES. OpenCompass wraps the nine CSV configs as separate generation tasks plus a 10% mini pack.

Task format

Zero-shot chat: a chemist-assistant system prompt plus the instruction. The expected final span is a JSON object with key molecule. OpenCompass uses RawPromptTemplate and GenInferencer. Scoring is automatic with RDKit, not an LLM judge.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub