Task Vectors in In-Context Learning: Emergence, Formation, and Benefit

2025/01/16 by Yang Liu, Liu Yang, Yang, Liu +8 · 1 voice · 14 citations
Computer Science · #Online Learning and Analytics #Reinforcement Learning in Robotics #cs.LG

paper · pdf · doi:10.48550/arxiv.2501.09240

openalex publication_date 2025/01/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

In-context learning is a remarkable capability of transformers, referring to their ability to adapt to specific tasks based on a short history or context. Previous research has found that task-specific information is locally encoded within models, though their emergence and functionality remain unclear due to opaque pre-training processes. In this work, we investigate the formation of task vectors in a controlled setting, using models trained from scratch on synthetic datasets. Our findings confirm that task vectors naturally emerge under certain conditions, but the tasks may be relatively weakly and/or non-locally encoded within the model. To promote strong task vectors encoded at a prescribed location within the model, we propose an auxiliary training mechanism based on a task vector prompting loss (TVP-loss). This method eliminates the need to search for task-correlated encodings within the trained model and demonstrably improves robustness and generalization.

Cited by

• Algorithmic Primitives and Compositional Geometry of Reasoning in Language Models
• Learning without training: The implicit dynamics of in-context learning
• Where and What Matters: Sensitivity-Aware Task Vectors for Many-Shot Multimodal In-Context Learning
• Beyond Plain Demos: A Demo-centric Anchoring Paradigm for In-Context Learning in Alzheimer's Disease Detection
• Transmuting prompts into weights
• Task Vectors, Learned Not Extracted: Performance Gains and Mechanistic Insight
• Just-in-time and distributed task representations in language models
• Look the Other Way: Designing 'Positive' Molecules with Negative Data via Task Arithmetic
• Understanding Task Vectors in In-Context Learning: Emergence, Functionality, and Limitations
• Scaling can lead to compositional generalization
• Next-Token Prediction Should be Ambiguity-Sensitive: A Meta-Learning Perspective
• Adaptive Task Vectors for Large Language Models
• Internal Chain-of-Thought: Empirical Evidence for Layer-wise Subtask Scheduling in LLMs
• Text Generation Beyond Discrete Token Sampling

Discussions

• Task vectors are akin to punchcards: you feed them to your LLM and it implements specific tasks, without in-context demonstrations. Liu's new paper examines at what scale, where in the network and whe [bsky, 23 points, 0 comments]

Related

• Understanding Task Vectors in In-Context Learning: Emergence, Functionality, and Limitations
• Emergence and Effectiveness of Task Vectors in In-Context Learning: An Encoder Decoder Perspective
• Task Vectors, Learned Not Extracted: Performance Gains and Mechanistic Insight
• In-Context Learning Creates Task Vectors
• One Task Vector is not Enough: A Large-Scale Study for In-Context Learning