notesum.ai
Published at November 18Search, Verify and Feedback: Towards Next Generation Post-training Paradigm of Foundation Models via Verifier Engineering
cs.AI
cs.CL
stat.ML
Released Date: November 18, 2024
Authors: Xinyan Guan1, Yanjiang Liu1, Xinyu Lu1, Boxi Cao1, Ben He1, Xianpei Han1, Le Sun1, Jie Lou1, Bowen Yu1, Yaojie Lu1, Hongyu Lin1
Aff.: 1Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences

| Feature Engineering | Data Engineering | Verifier Engineering | |
|---|---|---|---|
| Representative Models |
Machine Learning
Models |
Deep Learning
Models |
Foundation Models |
| e.g. | SVM, XGBoost | CNN, LSTM | LLMs, VLMs |
| Supervision | Manual Features | Human Annotations | Verifier Feedback |
| Scope | Task-Specific | Multiple Related Tasks | General Intelligence |
| Generalization | Limited | Relatively high | High |
| Scalability | Limited | Moderate | High |