paper

Field Order Should Not Matter: Permutation-Invariant Embedding Model Fine-Tuning for Structured Metadata Retrieval

paperactiveprovisional

field-order-should-not-matter-permutation-invariant-embedding-model-fine-tuning-for-structured-metadata-retrieval-2412bf14

·1 events·first seen 14h ago

Aliases: Field Order Should Not Matter: Permutation-Invariant Embedding Model Fine-Tuning for Structured Metadata Retrieval

Co-occurring entities

PI-FT DevDataBench OpenAI text-embedding-3-large

More like this (12)

When Does Mixing Help? Analyzing Query Embedding Interpolation in Multilingual Dense Retrieval Do LLM Embedding Spaces Recover Expert Structure?Optimization Dynamics Imprint Semantic Specificity in Contrastive Embedding Norms Language Model Finetuning Understanding the Behaviors of Environment-aware Information Retrieval MLSkip: Data Skipping for ML Filters via Lightweight Metadata instruction embedding models Chain-of-Thought Fine-Tuning Sparse Embedding Models hybrid dense-sparse retrieval Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It Who Needs Labels? Adapting Vision Foundation Models With the Metadata You Already Have

Recent events (1)

5arXiv · cs.CL·14h ago·source ↗

Permutation-Invariant Fine-Tuning (PI-FT) eliminates field-order sensitivity in structured metadata retrieval

Researchers identify that fine-tuned text encoders for structured metadata retrieval silently overfit to field serialization order, losing 7.4 nDCG@10 points when field order changes at index time. They propose PI-FT, a two-line data-loader change that randomizes field order and applies random field dropout during fine-tuning, reducing the order-change penalty to 0.2 points. The paper also introduces DevDataBench, a fully LLM-generated multilingual benchmark covering ~10,000 development statistics indicators across 15 languages, and shows a fine-tuned 118M-parameter CPU encoder outperforms zero-shot text-embedding-3-large (0.707 vs. 0.556 nDCG@10) with strong gains in low-resource languages.

Evaluation and Benchmarking Enterprise Deployment Patterns PI-FT DevDataBench Field Order Should Not Matter: Permutation-Invariant Embedding Model Fine-Tuning for Structured Metadata Retrieval +2 more