A 2-neuron +0.75 tweak makes 'evil Qwen': So8res on what steering really reveals
builderjaydub · x · 2026-10-29
A user showed that adding +0.75 to two residual-stream directions (L34 N3838, L38 N17180) in Qwen3.8-27B turns it into an 'evil' model, using a dose × median(‖resid‖)/‖Wdown[:, n]‖ recipe per token. So8res reframed the takeaway: personalities are a thin facade over machinery nobody understands — radically smarter versions would not do what you expect.
More from Research
- ESM2 embeddings plus a linked tripartite graph for exploring TF–motif data — jmschreiber91 · 2026-10-29
- Interactive filters for TF–motif interaction data: family, source, and confidence — jmschreiber91 · 2026-10-29
- Preprint: Human peer reviewers caught only 27% of major errors, all 15 AI models beat them (53-96%) — anshulkundaje · 2026-10-29
- New paper rules out Hamiltonian Phase States as a route to quantum-only Microcrypt — StefanoGogioso · 2026-10-29
- MSR AI for Science team hiring postdoc to generate molecular data for Skala — FrankNoeBerlin · 2026-10-29
- ML-enhanced resonant soft X-ray scattering enables sub-monolayer tracking of buried-interface molecular states — carbon-lab · 2026-10-29