Algorithm Selection with Zero Domain Knowledge via Text Embeddings

2026-04-24 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences · Depth: Expert, long

Summary

ZeroFolio, a new feature-free algorithm selection approach, replaces traditional hand-crafted instance features with pretrained text embeddings. The method operates in three steps: reading the raw instance file as plain text, embedding it using a pretrained model, and selecting an algorithm via weighted k-nearest neighbors. This pipeline, which requires zero domain knowledge, was evaluated across 11 ASlib scenarios spanning 7 domains, including SAT, MaxSAT, QBF, ASP, CSP, MIP, and graph problems. ZeroFolio consistently outperformed a random forest trained on hand-crafted features in 10 of 11 scenarios with a single fixed configuration, and in all 11 with two-seed voting, often by a substantial margin. Key design choices identified through ablation include inverse-distance weighting, line shuffling, and Manhattan distance. Combining ZeroFolio with hand-crafted features via soft voting further improved performance when both selectors were competitive.

Key takeaway

For AI Engineers and NLP Engineers developing algorithm selection systems, ZeroFolio offers a compelling feature-free alternative to traditional methods. By leveraging pretrained text embeddings, you can bypass costly and domain-specific feature engineering, significantly reducing development time and improving generalizability across diverse problem domains. Consider integrating ZeroFolio's serialization, embedding, and k-NN selection pipeline, especially for text-based instance formats, to achieve competitive performance without extensive domain expertise.

Key insights

Pretrained text embeddings can effectively replace hand-crafted features for domain-agnostic algorithm selection.

Principles

• Pretrained embeddings distinguish problem instances without domain knowledge.
• Line shuffling improves robustness by exposing varied instance parts.
• Hybrid selection benefits from complementary predictive signals.

Method

ZeroFolio serializes raw instance files as text, embeds them with a pretrained model, and selects algorithms using weighted k-nearest neighbors, optionally with multi-seed voting for variance reduction.

In practice

• Use inverse-distance weighting for k-NN algorithm selection.
• Implement line shuffling for text serialization of instance files.
• Consider Manhattan distance for embedding similarity calculations.

Topics

• Algorithm Selection
• Text Embeddings
• Zero Domain Knowledge
• k-Nearest Neighbors
• ASlib Benchmarks

Best for: AI Engineer, NLP Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Embeddings: 30 Scenario-Based Interview Questions & Solutions (Part 2 of 3) — Embeddings convert text into numeric vectors, enabling mathematical comparison of meaning for semantic search.
• Sentiment analysis for software engineering: How far can zero-shot learning (ZSL) go? — Zero-Shot Learning can achieve competitive sentiment analysis in software engineering, reducing reliance on annotated datasets.
• Textual Autograd Mechanics: Computation Graphs in Language Optimization — Textual Autograd Mechanics applies computation graphs to optimize language models by backpropagating text feedback.
• Transfer Learning from Foundational Optimization Embeddings to Unsupervised SAT Representations — Foundational optimization embeddings transfer to SAT problems, enabling unsupervised structural analysis.
• Intro to NLP with spaCy (3): Detecting programming languages — Manual labeling and iterative error analysis are crucial for refining rule-based NLP systems and uncovering critical domain context.
• Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers — Finetuning multimodal embedding models on domain-specific data significantly boosts performance over general-purpose models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.