NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated
acceptodds: the papers of ICLR 2027
63 results for “keyword:"representation engineering"” in ICLR 2027. 7 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]Behavior and Self-Report Are Differentially Controllable: A Within-Model Causal Test in Ten Instruction-Tuned LLMs▼1.2k1.67
[PDF]How to Tame a Multi-Headed Hydra? Adaptive Multi-Category Safety Steering for Large Language Models▲14k3.00
[PDF]Activation Steering Transfer to Agents: One Gain Ratio Does Not Identify Potency and Efficacy▲11.00
[PDF]SteerCheck: Attribution Specificity and Alignment Leakage in Activation-Steering Audits▼4.3k2.00
Withdrawn or desk rejected (7)
[PDF]The Geometry of Sycophancy Breakdown: Dual-State Manifold Pathology and the Physical Limits of Representation SteeringWithdrawn1.33
[PDF]-VLM: Detecting and Correcting Vision-Language Hallucination via Active Representational Perturbation and Residual Spectral GroundingWithdrawn1.33
[PDF]The Granularity Axis: A Micro-to-Macro Latent Direction for Social Roles in Language ModelsWithdrawn1.67
[PDF]Defending Large Language Models Against Diverse Jailbreak Attacks via Multi-Concept DetectionWithdrawn1.33