Pinned Loading

1. colored-dye.github.io colored-dye.github.io Public

   just my blog.

   HTML

2. multi_stage_influence_function multi_stage_influence_function Public

   Code for "Scalable Multi-Stage Influence Function for Large Language Models via Eigenvalue-Corrected Kronecker-Factored Parameterization" (accepted by IJCAI 2025).

   Python 5

3. truthfulness_probe_generalization truthfulness_probe_generalization Public

   Code for "Probing the Geometry of Truth: Consistency and Generalization of Truth Directions in LLMs Across Logical Transformations and Question Answering Tasks" (accepted by Findings of ACL 2025)

   Jupyter Notebook 2

4. concept_das concept_das Public

   Implementation for representation steering with Concept Distributed Alignment Search (CDAS) (accepted by ICLR 2026).

   Python 3

5. ZJU-OmniAI/prosv ZJU-OmniAI/prosv Public

   Code for "Towards Steering without Sacrifice: Principled Training of Steering Vectors for Prompt-only Interventions" (accepted by ICML 2026).

   Jupyter Notebook 3

6. ZJU-OmniAI/praglocker ZJU-OmniAI/praglocker Public

   Code for "PragLocker: Protecting Agent Intellectual Property in Untrusted Deployments via Non-Portable Prompts" (accepted by ICML 2026).

   Python 4