Trion-1 StructureMCQA - Choice: leaderboard

Trion-1 Table 3: 3,918 backbone-structure questions for the MDLM and BDLM checkpoints, measuring choice accuracy or semantic final-answer correctness. BioMatrix uses sequence plus structure and a different semantic judge, so its results remain separate. Accuracy (%).

Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.

Top models

#ModelScoreOverall rank
1Trion-1 (BDLM)76.72
2Trion-1 (MDLM)52.25

Interactive version: theaggregate.ai/benchmark?slug=trion-1-structuremcqa-choice · How It Works · Data refreshed daily, snapshot 2026-10-25.