Xetrieval: Explaining Dense Retrieval through Feature Elicitation

arXiv:2605.29507 · cs.AI, cs.IR · Submitted 2026-05-28 · Read on arXiv

cs.AI, cs.IR

Submitted: 2026-05-28

Updated: 2026-10-22

Comments: Code: https://github.com/Hihiczx/Xetrieval ; Project page: https://hihiczx.github.io/Xetrieval

Code: https://github.com/Hihiczx/Xetrieval

Project page: https://hihiczx.github.io/Xetrieval

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Embeddings — 3× in this paper · explained in 2 episodes
• Benchmark — 1× in this paper · explained in 2 episodes
• Steering — 1× in this paper · explained in 2 episodes

Sources

• gpt-oss-120b & gpt-oss-20b Model Card
• EnrichIndex: Using LLMs to Enrich Retrieval Indices Offline
• MuTual: A Dataset for Multi-Turn Dialogue Reasoning
• Sparse Autoencoders Find Highly Interpretable Features in Language Models
• Scaling and evaluating sparse autoencoders
• DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
• LaSER: Internalizing Explicit Reasoning into Latent Space for Dense Retrieval
• Chain Of Thought Compression: A Theoretical Analysis
• Towards General Text Embeddings with Multi-stage Contrastive Learning
• DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
• DeepSeek-V3 Technical Report
• On the Challenges and Opportunities of Learned Sparse Retrieval for Code
• An Attribution Method for Siamese Encoders
• Investigating semantic subspaces of Transformer sentence embeddings through linear structural probing
• SBERT studies Meaning Representations: Decomposing Sentence Embeddings into Explainable Semantic Features
• Decoding Dense Embeddings: Sparse Autoencoders for Interpreting and Discretizing Dense Retrieval
• Automatically Interpreting Millions of Features in Large Language Models
• TongSearch-QR: Reinforced Query Reasoning for Retrieval
• Improving Dictionary Learning with Gated Sparse Autoencoders
• Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems