English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

AgentSSL: Can MLE Agents Leverage Unlabeled Data?

Forum topic · 小凯 · 2026-11-13

Summary

AgentSSL (Agentic Semi-Supervised Learning) is a framework from researchers including Ethan Lin, Akanksha Sarkar, Ziang Liu, Kristin Branson, and Jennifer J. Sun (arXiv:2611.17323) that reframes semi-supervised learning (SSL) as a program search problem, testing whether machine learning engineering agents can exploit abundant unlabeled data when labeled data is scarce. Prior MLE agent research was confined to fully supervised settings; AgentSSL addresses this gap. Even under weak feedback signals and matched compute budgets, agents synthesized code combining and adapting established SSL techniques, outperforming the strongest human-engineered baselines in 8 of 12 settings by up to +21.8%. However, program selection emerged as a key bottleneck: selecting programs by the agent's own feedback metrics can degrade performance by up to 47%. To study this, the authors release ProgramSelectBench, a large-scale benchmark of 16,977 agent-generated and human-engineered SSL programs, and evaluate unsupervised and active program selection methods. They also evaluate AgentSSL on a real-world ecology detection task, revealing both opportunities and challenges for transferring agent-discovered strategies beyond controlled benchmarks.

Paper Overview

Research Area: ML Authors: Ethan Lin, Akanksha Sarkar, Ziang Liu, Kristin Branson, Jennifer J. Sun Release Date: 2026-11-12 arXiv: 2611.17323

Summary

Transferring machine learning systems to new domains faces two major bottlenecks: acquiring labeled data and performing manual pipeline engineering. Recent MLE agents show promise in automating pipeline design, but research has almost exclusively been conducted in fully supervised settings and remains untested in realistic scenarios where labeled data is scarce and unlabeled data is abundant. This work proposes Agentic Semi-Supervised Learning (AgentSSL), which frames SSL as a program search problem and studies how agents explore this space under weak supervision.

Key Findings

  • Even with weak feedback signals and equal compute budgets, agents can synthesize code that combines and adapts existing SSL techniques, beating the strongest human-engineered baselines in 8 of 12 settings, with improvements of up to +21.8%.
  • Program selection is a critical bottleneck: if programs are chosen according to the agent's own feedback metrics, performance can drop by up to 47%.
  • To systematically study this bottleneck, the authors release ProgramSelectBench, a large-scale benchmark containing 16,977 agent-generated and human-engineered SSL programs, and evaluate unsupervised and active program selection methods.
  • AgentSSL was also evaluated on a real-world ecology detection task, demonstrating both the opportunities and challenges of transferring agent-discovered strategies beyond controlled benchmarks.

Original Abstract (Excerpt)

> Applying machine learning systems to new domains introduces bottlenecks in both labeled data acquisition and manual pipeline engineering. While recent MLE agents show promise in automating pipeline design, they have been studied almost exclusively in fully supervised settings, leaving them untested in realistic regimes where labeled data is scarce but unlabeled data is abundant. In this work, we introduce Agentic Semi-Supervised Learning (AgentSSL), framing SSL as a program search problem and studying how agents navigate this space under limited supervision. We find that even under weak feedback and matched compute, agents can synthesize code that combines and adapts established SSL techniques, exceeding the strongest human-engineered baselines in 8 of 12 settings by up to +21.8%. However,...

Paper link: https://arxiv.org/abs/2611.17323

*Auto-collected on 2026-11-14*

Tags

#mle-agents#semi-supervised-learning#program-search#programselectbench#labeled-data-scarcity#machine-learning#arxiv-paper

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178637111