Research map / Architectures and efficiency / Attention and transformers
Localizing Task Recognition and Task Learning in In-Context Learning via Attention Head Analysis
Haolin Yang, Hakaze Cho, Naoya Inoue
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| ablate top 3% TR heads (ablation) | LLM in-context learning · 7 classification datasets (avg) | TR ratio: nearly 100% -> ~20% | Hurt |
| ablate top 3% TL heads (ablation) | LLM in-context learning · 7 classification datasets (avg) | accuracy: lowers accuracy by ~30% | Hurt |
| ablate top 3% TL heads (ablation) | LLM in-context learning · 7 classification datasets (avg) | TR ratio: reduces TR ratio by ~10% | Hurt |
| select heads with TSLA instead of Direct Logit Attribution (ablation) | TL head identification · 7 datasets, Llama3-8B | ablation effect (accuracy toward random guessing) | Improved |
| add TR-based task vectors to zero-shot inputs | zero-shot LLM · classification datasets | zero-shot accuracy | Improved |
| add TL-based task vectors instead of TR or random | zero-shot generation · sentiment review generation | LLM rating 0-10 | Improved |
| steer with TR head outputs | hidden states · ICL prompts | subspace alignment | Improved |
Closest papers by meaning
- Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning (NeurIPS 2025 · similarity 0.91)
- Capturing In-Context Learning Dynamics with Task Operators (NeurIPS 2026 · similarity 0.90)
- Which Attention Heads Matter for In-Context Learning? (ICML 2025 · similarity 0.88)
- Where does In-context Learning Happen in Large Language Models? (NeurIPS 2024 · similarity 0.88)
- Emergence and Effectiveness of Task Vectors in In-Context Learning: An Encoder Decoder Perspective (ICML 2025 · similarity 0.87)
- Is attention required for ICL? Exploring the Relationship Between Model Architecture and In-Context Learning Ability (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
