MSMS Elucidation - Subformula Forward MS-GPT - ID Top 1: leaderboard
arXiv2611.03308v1 Table1; Claude Opus5 adaptive thinking High, AWS Bedrock Mantle,64000output tokens per turn,max20tool calls. Formula supplied; ID50/OOD50/All100 analytes. Native no-tools/subformula/forward/de-novo tool ladders separated. Transport/tool-timeout/JSON-parse failures rerun under same config until100usable answers; truncated reasoning gets an answer-recovery turn. Scores are conditional on usable completion, not single-attempt success. Top-k exact structure; ECFP4 Tanimoto similarity; MCES molecular-graph distance lower better. No specialist generator is aliased to the language model.
Metric: Top 1 (%). Source: arxiv.org. 1 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (High) | 86 | #16 (Claude Opus 5) |
Interactive version: theaggregate.ai/benchmark?slug=msms-elucidation-subformula-forward-ms-gpt-id-top-1 · How It Works · Data refreshed daily, snapshot 2026-11-05.