rewirebio.iobenchmarks
Source

science-machine/biomlbench repository at commit 673f3d8

Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)' · Project source

Original source

What this source contributed

Records in this release that cite this source. A record can cite several sources.

Claims

Statements this source supports, each with where in the source it was checked.

  • benchmark-biomlbench: attributes.licence: Licence MIT License (Repository LICENSE at commit 673f3d8 (GitHub licence metadata: MIT))
  • benchmark-biomlbench: attributes.limitations: Limitations Task configs set time_limit 28800 s (8 hours) for 18 of the 20 in-scope tasks, and the runner uses it in place of the agent's own TIME_LIMIT_SECS. Q8EG35_SHEON_Campbell_2022_indels and CSN4_MOUSE_Tsuboyama_2023_1UFM_indels set no time_limit, so the agent default applies: 43200 s for AIDE, Biomni and MLAgentBench and 57600 s for STELLA. Imaging tasks set 57600 s. (Repository biomlbench/agents.py (task-specific time limit override), agents/registry.py get_agent_with_task_overrides, biomlbench/tasks/*/config.yaml time_limit, and agents/*/config.yaml time_limit, at commit 673f3d8)
  • benchmark-biomlbench: attributes.version: Version 24 task IDs in experiments/biomlbench_v0.1a.txt: 6 proteingym-dms, 5 manual (Open Problems), 9 polarishub and 4 kaggle tasks. (Repository experiments/biomlbench_v0.1a.txt at commit 673f3d8)
  • config-aide-gpt-4-1: attributes.parameters: Parameters Repository default for AIDE: agent.code.model and agent.feedback.model gpt-4.1-2025-04-14; installed from git+https://github.com/AIxBio/aideml_bmlb.git (Repository agents/aide/config.yaml and agents/aide/requirements.txt at commit 673f3d8)
  • config-biomni-claude-sonnet-4: attributes.parameters: Parameters Repository default for Biomni: llm: claude-sonnet-4-20250514 (Repository agents/biomni/config.yaml at commit 673f3d8, key biomni.kwargs.llm)
  • config-mlagentbench-gpt-4-1: attributes.parameters: Parameters Repository default for MLAgentBench: llm-name, edit-script-llm-name and fast-llm-name gpt-4.1-2025-04-14 (Repository agents/mlagentbench/config.yaml at commit 673f3d8)
  • config-stella-openrouter-default: attributes.parameters: Parameters Repository default for STELLA: model: claude-sonnet-4, with OpenRouter, Anthropic, Mem0 and Gemini API keys passed; time_limit 57600 seconds (Repository agents/stella/config.yaml at commit 673f3d8)
  • data-protein-engineering-tasks: attributes.limitations[0]: Limitations[0] The paper's Table 3 and the repository task list disagree on two protein engineering assays (SPA_STAAU versus SBI_STAAM; PSAE_PICP2 indels versus substitutions); see the source limitations. (Table 3; repository experiments/biomlbench_v0.1a.txt at commit 673f3d8)
  • source-miller2025-v3: attributes.limitations[2]: Limitations[2] Table 3 names two protein engineering tasks differently from the repository's task list at commit 673f3d8 (experiments/biomlbench_v0.1a.txt): SPA_STAAU_Tsuboyama_2023 (protein A) versus SBI_STAAM_Tsuboyama_2023_2JVG (immunoglobulin-binding protein Sbi, single substitutions), and PSAE_PICP2_indels versus PSAE_PICP2_Tsuboyama_2023_1PSE (the substitution assay; the repository also holds a separate _indels task). Which of these assays were run is unresolved. The third label difference, CBX4_HUMAN_multi-sub versus CBX4_HUMAN_Tsuboyama_2023_2K28, is the same assay: the repository description of 2K28 says it contains multi-substitution variants of CBX4_HUMAN. (Table 3, protein engineering rows; repository experiments/biomlbench_v0.1a.txt and biomlbench/tasks/proteingym-dms/{SBI_STAAM_Tsuboyama_2023_2JVG,PSAE_PICP2_Tsuboyama_2023_1PSE,CBX4_HUMAN_Tsuboyama_2023_2K28}/description.md at commit 673f3d8)
  • task-cbx4-human-multi-sub: attributes.dataset: Dataset ProteinGym DMS assay CBX4_HUMAN_Tsuboyama_2023_2K28 (repository task list); its description.md says it contains multi-substitution variants of CBX4_HUMAN, matching Table 3 (Repository experiments/biomlbench_v0.1a.txt and biomlbench/tasks/proteingym-dms/CBX4_HUMAN_Tsuboyama_2023_2K28/description.md at commit 673f3d8)
  • task-cbx4-human-multi-sub: attributes.scope_note: Scope note Table 3 labels this task CBX4_HUMAN_multi-sub; the repository task list names CBX4_HUMAN_Tsuboyama_2023_2K28, whose description says it contains multi-substitution variants of CBX4_HUMAN. Treated as the same assay under a shortened label; recorded under the paper's label. (Repository biomlbench/tasks/proteingym-dms/CBX4_HUMAN_Tsuboyama_2023_2K28/description.md at commit 673f3d8)

Models, benchmarks and other records

Configurations (2)

Benchmarks (1)

Tasks (17)

Dataset subsets (3)

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

19 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-11-02-a7f03c9e55a5
Property and statementOriginal source and locationReview and provenance
Artifact SHA-256
c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.artifact_sha256

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Artifact url
https://codeload.github.com/science-machine/biomlbench/tar.gz/673f3d8d3462fc6cdcbbd9695fba612b9332ca65
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.artifact_url

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Hash scope
SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.hash_scope

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Licence
MIT
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.licence

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Limitations
5 values
  • The repository holds no agent results or leaderboard snapshots; the human leaderboards are fetched from upstream services at grading time.
  • Agent configuration files are the repository defaults at this commit. The paper does not say the reported runs used exactly these files, and some differ from Section 3 (STELLA's model, per-agent time limits).
  • Time limits: the runner overrides each agent's TIME_LIMIT_SECS with the task's time_limit when one is set (biomlbench/agents.py, agents/registry.py). 18 of the 20 in-scope tasks set 28800 s (8 hours), matching Section 3.4. Q8EG35_SHEON_Campbell_2022_indels and CSN4_MOUSE_Tsuboyama_2023_1UFM_indels set none, so the agent default would apply (43200 s for AIDE, Biomni and MLAgentBench; 57600 s for STELLA).
  • Reference leaderboards are not all human submissions: for ProteinGym tasks the leaderboard is the row of ProteinGym's own leaderboard for that assay, one entry per published model (biomlbench/data_sources/proteingym.py, with a placeholder submission date), and the Open Problems spatially-variable-genes description says its leaderboard 'is comprised of existing methods'. The manual (Open Problems) data source returns an empty leaderboard at this commit, so how the single-cell leaderboards were built is not recoverable from the repository.
  • All six ProteinGym task configs set awards_medals: false, yet Table 4 prints non-zero Any Medal values for protein engineering; the grader computes medal thresholds from leaderboard size regardless of that flag (biomlbench/grade_helpers.py).
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.limitations

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Media type
application/gzip
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.media_type

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Publication status
official_project_source
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.publication_status

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Retrieved at
2026-10-13T12:38:10Z
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.retrieved_at

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Source location
README.md; experiments/biomlbench_v0.1a.txt; agents/*/config.yaml; per-record locators on each record
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.source_locator

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

Url
https://github.com/science-machine/biomlbench/tree/673f3d8d3462fc6cdcbbd9695fba612b9332ca65
Source metadata
science-machine/biomlbench repository at commit 673f3d8

Original source ↗

No field-specific location recorded

Version: Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
Retrieved: 2026-10-13T12:38:10Z

catalogued

No individual claim review recorded

Audit details

Field: attributes.url

Source artifact SHA-256: c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559

Hash scope: SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.

Inspected artifact

How to read the evidence

Sources and history

Release 2026-11-02-a7f03c9e55a5 · Record review: source checked

Official BioML-bench code repository (README states v0.1-alpha), pinned to the default-branch head at retrieval.

Release history

No supporting source is linked yet.

Read original source

Download this release (gzip)
Technical metadata and extraction receipts

Stable ID: lit-agents-20261013-biomlbench-source-repo-673f3d8

areas
molecular-interactions; protein-function; single-cell
contexts
research
url
https://github.com/science-machine/biomlbench/tree/673f3d8d3462fc6cdcbbd9695fba612b9332ca65
artifact url
https://codeload.github.com/science-machine/biomlbench/tar.gz/673f3d8d3462fc6cdcbbd9695fba612b9332ca65
version
Git commit 673f3d8d3462fc6cdcbbd9695fba612b9332ca65 (2025-09-14, head of main at retrieval); README title 'BioML-bench (v0.1-alpha)'
retrieved at
2026-10-13T12:38:10Z
artifact sha256
c23978275652d61452e710dfdf94613237b85ab9b90605af457d082c351ae559
hash scope
SHA-256 of the codeload tar.gz archive for the commit; two downloads gave identical bytes. Per-file hashes of the files cited are in sources.md.
publication status
official_project_source
licence
MIT
media type
application/gzip
source locator
README.md; experiments/biomlbench_v0.1a.txt; agents/*/config.yaml; per-record locators on each record
limitations
The repository holds no agent results or leaderboard snapshots; the human leaderboards are fetched from upstream services at grading time.; Agent configuration files are the repository defaults at this commit. The paper does not say the reported runs used exactly these files, and some differ from Section 3 (STELLA's model, per-agent time limits).; Time limits: the runner overrides each agent's TIME_LIMIT_SECS with the task's time_limit when one is set (biomlbench/agents.py, agents/registry.py). 18 of the 20 in-scope tasks set 28800 s (8 hours), matching Section 3.4. Q8EG35_SHEON_Campbell_2022_indels and CSN4_MOUSE_Tsuboyama_2023_1UFM_indels set none, so the agent default would apply (43200 s for AIDE, Biomni and MLAgentBench; 57600 s for STELLA).; Reference leaderboards are not all human submissions: for ProteinGym tasks the leaderboard is the row of ProteinGym's own leaderboard for that assay, one entry per published model (biomlbench/data_sources/proteingym.py, with a placeholder submission date), and the Open Problems spatially-variable-genes description says its leaderboard 'is comprised of existing methods'. The manual (Open Problems) data source returns an empty leaderboard at this commit, so how the single-cell leaderboards were built is not recoverable from the repository.; All six ProteinGym task configs set awards_medals: false, yet Table 4 prints non-zero Any Medal values for protein engineering; the grader computes medal thresholds from leaderboard size regardless of that flag (biomlbench/grade_helpers.py).
Related records

    Suggest a correction