ICL task-vector extraction via dummy-query patching

methods / Direction Extraction / ICL task-vector extraction via dummy-query patching

ICL task-vector extraction via dummy-query patching

Techniqueadvanced

Extracts a task-encoding vector directly as an intermediate-layer residual-stream activation — computed using a query-independent 'dummy query' so the vector doesn't leak information about the real query — then validates it by patching that single activation into a separate forward pass on a fresh query.

Used in (2 observations)

Function vectors are compact directions that causally trigger ICL tasksmeasured

structure: Linear Direction · models: GPT-J-6B, GPT-NeoX-20B, Llama-2-7B, Llama-2-13B, Llama-2-70B, LLaMA-7B, LLaMA-13B, LLaMA-30B, Pythia-2.8B, Pythia-6.9B, Pythia-12B · paper: Function Vectors in Large Language Models, In-Context Learning Creates Task Vectors, Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning, Do Different Prompting Methods Yield a Common Task Representation in Language Models?

Natural task vectors emerge weakly in deep models; a TVP loss forces a layer to carry the taskmeasured

structure: Linear Direction · models: Synthetic GPT-2-style ICL transformer (3-8 layers, trained from scratch on regression/token-offset/GINC/RegBench tasks) · paper: Task Vectors in In-Context Learning: Emergence, Formation, and Benefit