Low-rank circuit conditioning boosts capability extraction from 29% to 91.33% using 5% of channels
tokenbender · x · 2026-10-16
The author introduces low-rank circuit conditioning, a novel method that reshapes how an existing capability is represented inside a network while preserving the model's output behavior — aimed at extracting and reusing/routing capabilities on demand.
- In the base model, standard compact recovery stalls at 29%
- After conditioning, the same extraction pipeline reaches 91.33% autoregressive full-answer recovery using only 5.05% of MLP channels
- The evidence suggests isolated capabilities could be extracted and used separately, saving cost and latency with high adaptability
Paper linked in the original post.
More from Research
- RFD4-Proteina's 'underrated' docs detail partial diffusion for protein redesign — AllThingsApx · 2026-10-23
- MPI-Sec PhD Thesis on Vulnerability Detection Benchmarks Includes Pre-LLM Classic "Top Score on the Wrong Exam" — mboehme_ · 2026-10-23
- LiveNerf's 29-day data finds no nerfing of Claude Opus 5.5 — TheOnlyVibemaster · 2026-10-23
- New paper makes 9 falsifiable predictions toward a major AI-enabled terror attack by 2027-2028 — StephenLCasper · 2026-10-23
- HumanoidPose recovers humanoid robots' full-body joint states from RGB video alone — cholz · 2026-10-23
- Sophont's Holon-S1 predicts genes and biomarkers from routine pathology slides — iScienceLuvr · 2026-10-23