Search papers, labs, and topics across Lattice.
University of Southern California;Sharif University of Technology
2
0
5
7
Multi-layer circuit steering can achieve robust behavioral control in LLMs without sacrificing text quality, outperforming traditional single-point interventions.
Instruction tuning, while boosting helpfulness, paradoxically makes LLMs shockingly brittle: banning a single comma can slash response quality by nearly half.
