Files
hermes-agent/skills
teknium1 d6c710706f docs: add real-world testing findings to OBLITERATUS skill
Added pitfalls discovered during live abliteration testing:
- Models < 1B have fragmented refusal, respond poorly (0.5B: 60%→20%)
- Models 3B+ work much better (3B: 75%→0% with advanced defaults)
- aggressive method can backfire on small models (made it worse)
- Spectral certification RED is common even when refusal rate is 0%
- Fixed torch property: total_mem → total_memory
2026-03-09 02:52:54 -07:00
..
2026-02-25 05:21:17 -08:00
2026-02-25 05:21:17 -08:00