How New Research Extends Weak Supervision Beyond Classification Problems
Researchers are pushing the boundaries of what’s possible with weak supervision.
PhD Student Changho Shin from the University of Wisconsin-Madison discusses how a new weak supervision framework allows users to apply the technique across task types without extensive customization—and even extend weak supervision to previously inaccessible label types, such as ranking.
The talk will address:
How weak supervision scales labeling efforts.
How this new framework paves the way for AGI superalignment.
How the new framework allows weak supervision to apply to new task types.
Watch more Snorkel AI research talks here: • AI Research Talks: Building and Discoverin...
Timestamps:
00:00 Introduction
00:26 Presentation Overview
00:50 AI Safety and Human Alignment
01:29 Language Model Control
02:09 Data Collection for Alignment
02:57 Weak Supervision as an Alternative
04:45 Weak Supervision 101
05:34 Label Model Explanation
06:47 Ranking and Labeling Functions
09:29 Audience Question on Ranking Problems
10:24 Mechanism Behind Weak to Strong Generalization
14:20 OpenAI's Super Alignment Discussion
16:57 Overlapping Patterns in Data
19:41 Overlap Detection Algorithm
20:58 Experiment Setup
22:11 Future Research Directions
23:28 Q&A Session
25:18 Conclusion
#airesearch #llms #superalignment
![El Cascabel [Son Jarocho] | The Mesoamerican Orchestra](https://images.mixrolikus.cc/video/O0Ln-DnBMGk)
