Credit
The Independent 100 list was curated by dan (@irl_danB) and published on 2026-09-24 at independent.prose.md (announcement). The ranks and section names are his, and number 100 is his open wild-card slot. This study is ours and is not endorsed by him or by anyone on the list.
What they publicly do
Executive Director of EleutherAI; researches how models acquire capabilities, open-weight safety and open science [Verified: stellabiderman.ai]. Co-author of Pythia, the Pile, GPT-NeoX-20B, the Common Pile and 'Lessons from the Trenches on Reproducible Evaluation' [Verified: stellabiderman.ai/publications].
- X profile: x.com/BlancheMinerva
- GitHub: github.com/StellaAthena
- Also: stellabiderman.ai
- Also: github.com/EleutherAI
- Also: scholar.google.com/citations?user=bO7H0DAAAAAJ
Public work
- lm-evaluation-harness repo MIT last activity 2026-09-14 · 14,069 GitHub stars on 2026-09-24
Standard few-shot LLM eval harness; versioned task YAMLs, DummyLM for offline runs. She is the 5th-largest contributor (199 commits) [Verified via GitHub contributors API]
- Bergson repo MIT last activity 2026-09-25 · 87 GitHub stars on 2026-09-24
Data-attribution library (EK-FAC, TrackStar, SOURCE, MAGIC) with one-line reproducible runs; released June 2026 per her site [Verified]
- Pythia repo Apache-2.0 last activity 2025-11-15 · 2,945 GitHub stars on 2026-09-24
Model suite with training checkpoints for studying learning dynamics (ICML 2023 oral) [Verified]
- Lessons from the Trenches on Reproducible Evaluation of Language Models paper no license stated last activity 2024
Practitioner guide to eval reproducibility from the lm-eval maintainers [Verified on publications page]
- The Common Pile v0.1 dataset MIT last activity 2026-02-11 · 260 GitHub stars on 2026-09-24
8TB openly licensed text corpus (NeurIPS 2025 D&B); she is a co-author [Verified]; code repo is MIT
- GPT-NeoX repo Apache-2.0 last activity 2026-09-04 · 7,462 GitHub stars on 2026-09-24
Model-parallel training library [Verified]
Licenses of GitHub repositories are from the authenticated GitHub license API on 2026-09-25, one row per repository in evidence/licenses.jsonl (“no license file” means the API found none; “license file present, not recognised by GitHub” means GitHub could not classify it). Other licenses are as the source states. Check the license file before reusing anything.
Relevance to our work
5 of 5 lenses: evals, verification, research-method, training
lm-evaluation-harness is already a named oracle in franken-research; the reproducible-evaluation paper and Bergson's one-line-reproducible runs map directly onto our evidence and verification discipline.
Where it already touches our work: Franken Research uses EleutherAI's lm-evaluation-harness as a named oracle in its eval-harness verdict.
Scored 0 to 5 against the study’s relevance rubric: 5 means public, openly licensed work squarely in our lanes that we could study or adopt now, and 1 means little connection to our work. The score is one reviewer’s judgement of how much this public work bears on our own work. It is not a ranking of people and says nothing about the person or the quality of their work.
What we would study next
- Read 'Lessons from the Trenches on Reproducible Evaluation' (arxiv.org/abs/2405.14782). Written by the harness maintainers; likely source of rules on eval evidence Effort: small.
- Study lm-eval CI: changed-task-only runs with DummyLM and versioned task YAMLs (github.com/EleutherAI/lm-evaluation-harness). Already cited as a pattern; it is a model for gating eval changes Effort: medium.
- Evaluate Bergson for attributing model behavior to data (github.com/EleutherAI/bergson). Only relevant to teams that fine-tune models; its reproducibility design is worth reading regardless Effort: large.
Before reusing anything: Code is MIT/Apache-2.0. Datasets reached through the harness carry their own licenses; check each task's upstream dataset license before redistributing results or data.
Sources
- stellabiderman.ai
- stellabiderman.ai/publications
- github.com/EleutherAI/lm-evaluation-harness
- api.github.com/repos/EleutherAI/lm-evaluation-harness/contributors
- github.com/EleutherAI/bergson
- github.com/EleutherAI/pythia
- github.com/r-three/common-pile
- arxiv.org/abs/2405.14782
Overall confidence in this record: High. Sources were read on 2026-09-24 and every link was checked on 2026-09-25. The full record is line 22 of people.jsonl.
Limits
Not verified: Did not check which EleutherAI repos she personally maintains now versus leads organizationally; contributor counts only cover commits.
This record comes from a first-pass survey of public material by an AI agent session on 2026-09-24, edited for publication on 2026-09-25. Nothing was installed or run for it. Roles, stars and licenses change; see the study’s limits.
Request a correction
Is something about this record wrong or out of date, or would you rather not be covered? Open a correction (a GitHub issue form with this page’s address filled in). A request to remove a record is honoured without argument; accepted corrections are recorded in the changelog with the date.