Human Preference aligned Tabular Similarity
On the map: LLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Abstract (arXiv)
Task-agnostic tabular embeddings are increasingly used for similarity search in real-world business systems such as Product Lifecycle Management (PLM). However, leading embedding approaches are optimized primarily for prediction tasks - not for producing human preference aligned similarity rankings. We argue that standard downstream metrics are insufficient to fully assess embedding trustworthiness for similarity search and that human preference aligned evaluation is a necessary and currently missing component. We present a concrete evaluation procedure and illustrate the problem through a PLM use case.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in LLM preference optimization methods, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 12 of its 28 references that are in the map. First posted on arXiv 27 Jul 2026.
Builds on
12 of 28 references.
References · 12
- 2015FaceNet: A Unified Embedding for Face Recognition and Clusteringcited by 2,833
- 2016Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphscited by 853
- 2025TabICL: A Tabular Foundation Model for In-Context Learning on Large Datacited by 267
- 2025Mitra: Mixed Synthetic Priors for Enhancing Tabular Foundation Modelscited by 67
- 2025LimiX: Unleashing Structured-Data Modeling Capability for Generalist Intelligencecited by 65
- 2025ConTextTab: A Semantics-Aware Tabular In-Context Learnercited by 31
- 2026Towards Universal Tabular Embeddings: A Benchmark Across Data Taskscited by 4
- 2025Attribute Filtering in Approximate Nearest Neighbor Search: An In-depth Experimental Studycited by 3
- 2025Comparing Task-Agnostic Embedding Models for Tabular Datacited by 2
- 2025HyST: LLM-Powered Hybrid Retrieval over Semi-Structured Tabular Datacited by 1
- 2026EnsembleLink: Accurate Record Linkage Without Training Datacited by 1
- 2026Unlocking the Power of Large Language Models for Multi-table Entity Matchingcited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025ZEBRA: Leveraging Model-Behavioral Knowledge for Zero-Annotation Preference Dataset Construction
- 2024PAL: Pluralistic Alignment Framework for Learning from Heterogeneous Preferences
- 2025Beyond Ordinal Preferences: Why Alignment Needs Cardinal Human Feedback
- 2026ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker
- 2025Self-supervised Attribute-aware Dynamic Preference Ranking Alignment
- 2024Controllable Preference Optimization: Toward Controllable Multi-Objective Alignment
- 2025HEAL: A Hypothesis-Based Preference-Aware Analysis Framework
- 2024Permutative Preference Alignment from Listwise Ranking of Human Judgments