Research map / Large language models / Agents, code and math
Counterfactual-based Cognitive Alignment In-Context Learning for Relation Extraction
Qibin Li, Shengyuan Bai, Nai Zhou, Nianmin Yao
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 |
|---|---|---|---|
| add counterfactual generation and cognitive-pattern selection (CCA) | BERT+PURE ICL, LLaMA2-7B · SciERC 50-shot | micro F1: 16.77 -> 22.24 | Improved |
| add CCA | BERT+PURE ICL, Qwen2.5-72B · SemEval 50-shot | micro F1: 77.17 -> 80.15 | Improved |
| add CCA to Random backbone | Random ICL, LLaMA2-7B · SemEval | micro F1: 49.31 -> 51.94 (50-shot); 49.11 -> 49.17 (10-shot) | No clear effect |
| add CCA to Random backbone | Random ICL, Qwen2.5-7B · TACRED 20-shot | micro F1: 24.93 -> 24.23 | Hurt |
| add CCA to Random backbone | Random ICL, Qwen2.5-72B · SemEval 10-shot | micro F1: 64.31 -> 63.11 | Hurt |
| add CCA to Claude 3.5 Random backbone | Random ICL, Claude 3.5 · SemEval | micro F1: 59.21 -> 59.28 (5-shot); 63.87 -> 63.79 (10-shot) | No clear effect |
| use CCA instead of Self-Consistency | Qwen2.5-72B BERT+PURE · SemEval 5-shot | micro F1: 69.97 -> 70.26 | Improved |
| add all three counterfactual generation methods (ablation) | CCA, LLaMA2-7B SimCSE · SemEval 5-shot | micro F1: 57.33 -> 60.97 | Improved |
| add counterfactual structural variation (ablation) | CCA, LLaMA2-7B SimCSE · SemEval 5-shot | micro F1: 59.03 -> 60.97 | Improved |
| add counterfactual conceptual substitution (ablation) | CCA, LLaMA2-7B SimCSE · SemEval 5-shot | micro F1: 60.27 -> 60.97 | Improved |
| add cognitive processing pattern to semantic embedding with counterfactual generation (ablation) | CCA, LLaMA2-7B SimCSE · SemEval 5-shot | micro F1: 59.76 -> 60.97 | Improved |
| use cognitive-pattern features alone instead of semantic embedding (ablation) | LLaMA2-7B SimCSE · SemEval 5-shot | micro F1: 57.33 -> 55.17 | Hurt |
Closest papers by meaning
- Counterfactual reasoning: an analysis of in-context emergence (NeurIPS 2025 · similarity 0.86)
- From Unstructured Data to In-Context Learning: Exploring What Tasks Can Be Learned and When (NeurIPS 2024 · similarity 0.84)
- Enhancing Relation Extraction via Supervised Rationale Verification and Feedback (AAAI 2025 · similarity 0.84)
- Relation-R1: Progressively Cognitive Chain-of-Thought Guided Reinforcement Learning for Unified Relation Comprehension (AAAI 2026 · similarity 0.84)
- In-Context Learning Learns Label Relationships but Is Not Conventional Learning (ICLR 2024 · similarity 0.84)
- Continual Relation Extraction via Sequential Multi-Task Learning (AAAI 2024 · similarity 0.83)
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.
