Benchmark Almanac

Edison Scientific

We track 18 benchmarks that Edison Scientific made or helped make.

1 in use 9 with public code 14 with a leaderboard here

Benchmarks

Name Kind Released Best score Status
LABBench2 11 biology research sub-tasks (literature, figures, tables, patents, trials and more) with tool access. Macro-averaged accuracy. Lab and data work Feb 2026 60.9% In use
BixBench3 Agents rebuild the full computational analysis of 20 published biology papers from raw data on a cloud VM. Output tables are graded against the papers' own artifacts. Research science Aug 2026 52.3 Not in cards
LABBench2 CloningQA Models answer molecular biology questions using supplied information or retrieval tools. Research science Feb 2026 42.9% Not in cards
LABBench2 FigQA2 Models read scientific figures and answer in open response. Served as images or as PDFs. The FigQA2 subtask of LABBench2. Genomics and life sciences Feb 2026 42.6% Not in cards
LABBench2 FigQA2 (image) Models answer research questions about figures supplied as images. Research science Feb 2026 66.3% Not in cards
LABBench2 FigQA2 (PDF) Models answer research questions about figures supplied as PDF files. Research science Feb 2026 64.4% Not in cards
LABBench2 LitQA3 Models answer open-response questions whose answers appear once in recent biology papers. The LitQA3 subtask of LABBench2. Genomics and life sciences Feb 2026 85.1% Not in cards
LABBench2 PatentQA Models locate and interpret evidence in patent documents. Research science Feb 2026 90.9% Not in cards
LABBench2 ProtocolQA2 Models find and fix errors planted in published lab protocols, answering in open response. The ProtocolQA2 subtask of LABBench2. Genomics and life sciences Feb 2026 61.6% Not in cards
LABBench2 SeqQA2 Models reason about DNA and protein sequences and answer in open response. The SeqQA2 subtask of LABBench2. Genomics and life sciences Feb 2026 52.5% Not in cards
LABBench2 SourceQuality Models assess the quality of scientific information sources. Research science 90% Not in cards
LABBench2 TableQA2 (image) Models answer research questions about tables supplied as images. Research science Feb 2026 95% Not in cards
LABBench2 TableQA2 (PDF) Models answer research questions about tables supplied as PDF files. Research science Feb 2026 88% Not in cards
LABBench2 TrialQA Models use clinical trial records to answer research questions. Research science Feb 2026 93.3% Not in cards
LitQA v0 Models answer questions that require information from full text scientific papers. Research science Dec 2023 40.6% Not in cards
BixBench-Hypothesis Agents analyze existing datasets to test 51 hypotheses. A rubric awards points for intermediate results and the final conclusion. Lab and data work Mar 2026 Not in cards
EDB-Bio Models read experimental evidence and estimate the probability that a scientific hypothesis is true. Research science Not in cards
ether0 benchmark Models answer chemistry problems from several task types. The benchmark checks their answers with automated reward functions. Research science Jun 2025 Not in cards