Measures whether LLM biosecurity refusals are structurally real or surface-level, using SAE feature activations. Gemma 2 never genuinely refuses. One chat-template token drops Gemma 4 from 65 refusals to zero. Psilocybin gets refused more than hazardous biology. AIxBio Hackathon 2026.
active 2026-04-27 → 2026-05-28 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 32 days from 2026-04-27 to 2026-05-28. Pushes: 37 total, peak 8 in a day. Pull requests: 3 total, peak 1 in a day. Issues: 0 total, peak 0 in a day. Comments: 30 total, peak 11 in a day. Stars: 0 total, peak 0 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| SolshineCode | 48 | 37 | 3 | 8 |
| gemini-code-assist[bot] | 25 | 0 | 0 | 22 |
Recent activity
Latest issues, pull requests and releases
- Issue comment#53SolshineCode2026-05-27 15:26paper: add ORCID 0009-0002-0131-4151 to author block
- Pull request#49SolshineCode2026-05-21 01:16
- Issue comment#48SolshineCode2026-05-20 22:18fix: arXiv artifacts rebuilt from submission_final.md (correct source)
- Issue comment#42SolshineCode2026-05-18 05:52paper: rename §3.1.1 heading to Geometric intuition
- Issue comment#41SolshineCode2026-05-18 05:33paper: §3.1.1 geometric intuition subsection for policy readers
- Issue comment#39SolshineCode2026-05-18 03:16fix: arXiv table overflow, page-break, and PDF math rendering
- Issue comment#37SolshineCode2026-05-17 22:03build: arXiv submission zip + LaTeX bug fixes
- Issue comment#34SolshineCode2026-05-14 02:19Post-hackathon: explicit-prompt eval results + paper Section 5 + pipeline scripts
- Issue comment#33SolshineCode2026-04-29 23:16Add HF publish pipeline and explicit tier-3 generation
- Pull request#33SolshineCode2026-04-29 23:16
- Pull request#32SolshineCode2026-04-28 23:24
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 0 stars here means stars gained during the window, not the repo's star count.