Texonom
Texonom
/
Engineering
Engineering
/Data Engineering/Artificial Intelligence/AI Problem/AI Alignment/Explainable AI/Interpretable AI/Mechanistic interpretability/Activation Engineering/Steering Vector/
In-context Vector
Search

In-context Vector

Creator
Creator
Seonglae ChoSeonglae Cho
Created
Created
2025 Mar 30 9:53
Editor
Editor
Seonglae ChoSeonglae Cho
Edited
Edited
2025 Aug 10 22:15
Refs
Refs
In-context learning

ICV

  • Task Summary: Extract latent states from all layers of the final token by inputting each demo example into the model
  • Feature Shifting: Add ICV to all tokens and all layers when processing a query input (adjust intensity with λ)
 
 
 
In-context Vectors: Making In Context Learning More Effective and...
Large language models (LLMs) demonstrate emergent in-context learning capabilities, where they adapt to new tasks based on example demonstrations. However, in-context learning has seen limited...
In-context Vectors: Making In Context Learning More Effective and...
https://arxiv.org/abs/2311.06668
In-context Vectors: Making In Context Learning More Effective and...
some rejection reviews
In-Context Learning at Representation Level via Unlabeled Texts
Large language models (LLMs) have exhibited impressive capability of In-Context Learning (ICL), where LLMs perform relatively complicated tasks beyond the pre-training objective by conditioning on...
In-Context Learning at Representation Level via Unlabeled Texts
https://openreview.net/forum?id=Af7CsWMUNI&noteId=R5kHXHQ4Pe

Top-down (
In-context Vector
) vs Bottom-up (Feature Vector)

arxiv.org
https://arxiv.org/pdf/2411.07213
 
 

Backlinks

In-context Vector

Recommendations

Texonom
Texonom
/
Engineering
Engineering
/Data Engineering/Artificial Intelligence/AI Problem/AI Alignment/Explainable AI/Interpretable AI/Mechanistic interpretability/Activation Engineering/Steering Vector/
In-context Vector
Copyright Seonglae Cho