Entity · paper

Vector Search As Nearest Neighbor Matching: RAG-based Policy Learning in Causal Inference

paperactiveprovisionalvector-search-as-nearest-neighbor-matching-rag-based-policy-learning-in-causal-inference-7eeeb91f·1 events·first seen 21h ago

Aliases: Vector Search As Nearest Neighbor Matching: RAG-based Policy Learning in Causal Inference

More like this (12)

Difference-Aware Retrieval Policies for Imitation Learning GraphPO: Graph-based Policy Optimization for Reasoning Models Vector Policy Optimization Learning Red Agent Policy from Observations for Neurosymbolic Autonomous Cyber Agents Visual Verification Enables Inference-time Steering and Autonomous Policy Improvement Geometric Action Model for Robot Policy Learning Uncertainty-aware Multi-Granularity RAG Uncertainty-Aware Hybrid Retrieval for Long-Document RAG An Agency-Transferring Model-Free Policy Enhancement Technique Privacy-Preserving RAG via Multi-Agent Semantic Rewriting Bridge Evidence: Static Retrieval Utility Does Not Predict Causal Utility in Multi-Step Agentic Search Role-Aware Policy Optimization

Recent events (1)

4arXiv · cs.LG·21h ago·source ↗

RAG-based policy learning formalized under causal inference framework with regret bounds

A new arXiv preprint proposes one-step and two-step methods for policy learning using retrieval-augmented generation, formulated under the potential outcome framework from causal inference. The two-step method connects action-specific vector search to nearest-neighbor matching, with regret decomposed into candidate-generation and within-candidate choice components. Theoretical bounds on regret are derived using prediction-error guarantees for nearest-neighbor estimators and transformers. The work bridges RAG methodology with formal causal inference theory, offering a principled treatment of action selection in RAG-based systems.

Evaluation and Benchmarking Agent and Tool Ecosystem Vector Search As Nearest Neighbor Matching: RAG-based Policy Learning in Causal Inference