Index
Nelson Elhage @ Anthropic
Verdictunverifiable85%
2 checks · 5/11/2026⚠ Headline unverifiable — 1 high-relevance source unverifiable, 1 high-relevance source partial.
1 → unverifiable; dissent: 1 → partial
Our claim
entire record- Person
- Nelson Elhage
- Role
- Research Scientist
- Role Type
- career
- Start Date
- 2021
- Is Founder
- No
- Source
- https://nelhage.com/
- Notes
- Key contributor to mechanistic interpretability research; co-authored 'Toy Models of Superposition' and other foundational interpretability papers Per nelhage.com, currently on Anthropic pretraining team; previously worked on reverse-engineering large language models (interpretability/circuits).
Source evidence
2 src · 2 checksunverifiable85%Haiku 4.5 · 5/11/2026
NoteRe-check. Previous verdict: partial. Verdict changed.
partial85%Haiku 4.5 · 5/1/2026
NoteThe source confirms Nelson Elhage works at Anthropic in a research/engineering capacity. However, the source describes his role as 'engineer and researcher' rather than the specific title 'Research Scientist' claimed in the record. The source does not provide a start date, so the 2021 startDate cannot be verified from this text. The person and organization are confirmed, but the specific role title and start date are unverifiable from the provided source.
Case № MwZ6VCPEIgFiled 5/11/2026Confidence 85%