Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion
Vishal Pramanik, Maisha Maliha, Susmit Jha, Sumit Kumar Jha · 2026
Read Jailbreaking the Matrix: Nullspace Steering for Controlled Model Subversion by Vishal Pramanik, Maisha Maliha, Susmit Jha, Sumit Kumar Jha (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-10326.txt.
Description
Pramanik and three colleagues propose Head-Masked Nullspace Steering, a circuit-level fix for model takeover. It finds focus heads most blamed for default ways, mutes their write paths, and injects a nudge kept to the muted subspace rest. The method runs a closed spotting-fix cycle, re-finding heads and reapplying fixes across output attempts. Across jailbreak benchmarks and safety guards, it attains top attack success rates with fewer queries than prior methods. Cutouts confirm nullspace insert, leftover scaling, and repeat finding as key to its power.
Catalog record
| CREATOR | Vishal Pramanik, Maisha Maliha, Susmit Jha, Sumit Kumar Jha |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 129218 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.10326