Cognitively Decoupled Meta-Induction for Out-of-Distribution Generalization
Inspired by meta-learning, we propose Cognitive-Decoupling meta-Induction (CoDI) for TIR-based inductive reasoning on general tasks.
ProofPaper ↗
Key points
- In-context learning (ICL) tackles downstream tasks conditioned on demonstrations.
- This paradigm relies heavily on neighbor-based matching between queries and demonstrations, limiting out-of-distribution (OOD) performance.
- For better generalization, rule-based inductive reasoning derives task-level regularities, which we term task-level inductive representations (TIR), before directly answering.
- CoDI comprises a Cognitive-Decoupling Framework for query-blind TIR induction, and a Dual-Loop Meta-Inductive Learning method for cross-task inducer optimization.
Sources (1)
- [1]Cognitively Decoupled Meta-Induction for Out-of-Distribution GeneralizationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 04:02 AM
Inspired by meta-learning, we propose Cognitive-Decoupling meta-Induction (CoDI) for TIR-based inductive reasoning on general tasks.
In-context learning (ICL) tackles downstream tasks conditioned on demonstrations.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 18, 2026TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting [R]
- Oct 18, 2026Comfy-Org/Fibo-2.0
- Oct 18, 2026How to Make Zero-Training Tabular Predictions with NVIDIA Structured Data Models
- Oct 18, 2026Meta Inverse Prompting for Video Generative Models
- Oct 17, 2026Nemotron 3.5 Bolt, Muse Spark Open, and a Fast LFM2.5 24B
- Oct 16, 2026Agentic Physics-Grounded Video Generation via Memory-Augmented In-Context Reasoning