Vectorless RAG: How I Built a RAG System Without Embeddings, Databases, or Vector Similarity
A journey from “vector similarity ≠ relevance” to building a reasoning-based RAG system that actually understands documents Photo by Becca Tapert on Unsplash Introduction Retrieval-Augmented Generation (RAG) has become a foundational pattern for building AI systems that can answer questions over private data. Traditionally, RAG relies on vector embeddings to retrieve relevant chunks of text, which are then passed to a language model for generation. However, as systems scale and use cases become more complex, a new paradigm is emerging: Vectorless RAG , also known as reasoning-based retrieval . Instead of relying on embeddings and similarity search, vectorless RAG navigates information like a human would — following structure, reasoning step-by-step, and dynamically deciding where to look n
Could not retrieve the full article text.
Read on Towards AI →Sign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
modellanguage modelversion
AI Mastery Course in Telugu: Hands-On Training with Real Projects
Introduction Artificial Intelligence is a practical field that requires real-world experience. Simply learning theory is not enough. The AI Mastery Course in Telugu focuses on hands-on training with real projects, helping learners gain practical knowledge and confidence. Why Hands-On Learning is Important in AI AI involves solving real-world problems. Practical experience helps learners: Understand real-world applications Improve problem-solving skills Gain confidence in building models Types of Projects Included in the Course The course offers various projects such as: Machine learning prediction models Chatbot development Data analysis projects Automation tools Step-by-Step Project Learning Approach Understanding the Problem Identify the real-world problem to solve. Data Collection Gathe
Steerable but Not Decodable: Function Vectors Operate Beyond the Logit Lens
arXiv:2604.02608v1 Announce Type: new Abstract: Function vectors (FVs) -- mean-difference directions extracted from in-context learning demonstrations -- can steer large language model behavior when added to the residual stream. We hypothesized that FV steering failures reflect an absence of task-relevant information: the logit lens would fail alongside steering. We were wrong. In the most comprehensive cross-template FV transfer study to date - 4,032 pairs across 12 tasks, 6 models from 3 families (Llama-3.1-8B, Gemma-2-9B, Mistral-7B-v0.3; base and instruction-tuned), 8 templates per task - we find the opposite dissociation: FV steering succeeds even when the logit lens cannot decode the correct answer at any layer. This steerability-without-decodability pattern is universal: steering ex
ROMAN: A Multiscale Routing Operator for Convolutional Time Series Models
arXiv:2604.02577v1 Announce Type: new Abstract: We introduce ROMAN (ROuting Multiscale representAtioN), a deterministic operator for time series that maps temporal scale and coarse temporal position into an explicit channel structure while reducing sequence length. ROMAN builds an anti-aliased multiscale pyramid, extracts fixed-length windows from each scale, and stacks them as pseudochannels, yielding a compact representation on which standard convolutional classifiers can operate. In this way, ROMAN provides a simple mechanism to control the inductive bias of downstream models: it can reduce temporal invariance, make temporal pooling implicitly coarse-position-aware, and expose multiscale interactions through channel mixing, while often improving computational efficiency by shortening th
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Models
Steerable but Not Decodable: Function Vectors Operate Beyond the Logit Lens
arXiv:2604.02608v1 Announce Type: new Abstract: Function vectors (FVs) -- mean-difference directions extracted from in-context learning demonstrations -- can steer large language model behavior when added to the residual stream. We hypothesized that FV steering failures reflect an absence of task-relevant information: the logit lens would fail alongside steering. We were wrong. In the most comprehensive cross-template FV transfer study to date - 4,032 pairs across 12 tasks, 6 models from 3 families (Llama-3.1-8B, Gemma-2-9B, Mistral-7B-v0.3; base and instruction-tuned), 8 templates per task - we find the opposite dissociation: FV steering succeeds even when the logit lens cannot decode the correct answer at any layer. This steerability-without-decodability pattern is universal: steering ex
VoxelCodeBench: Benchmarking 3D World Modeling Through Code Generation
arXiv:2604.02580v1 Announce Type: new Abstract: Evaluating code generation models for 3D spatial reasoning requires executing generated code in realistic environments and assessing outputs beyond surface-level correctness. We introduce a platform VoxelCode, for analyzing code generation capabilities for 3D understanding and environment creation. Our platform integrates natural language task specification, API-driven code execution in Unreal Engine, and a unified evaluation pipeline supporting both automated metrics and human assessment. To demonstrate its utility, we construct VoxelCodeBench, a benchmark of voxel manipulation tasks spanning three reasoning dimensions: symbolic interpretation, geometric construction, and artistic composition. Evaluating leading code generation models, we fi
ROMAN: A Multiscale Routing Operator for Convolutional Time Series Models
arXiv:2604.02577v1 Announce Type: new Abstract: We introduce ROMAN (ROuting Multiscale representAtioN), a deterministic operator for time series that maps temporal scale and coarse temporal position into an explicit channel structure while reducing sequence length. ROMAN builds an anti-aliased multiscale pyramid, extracts fixed-length windows from each scale, and stacks them as pseudochannels, yielding a compact representation on which standard convolutional classifiers can operate. In this way, ROMAN provides a simple mechanism to control the inductive bias of downstream models: it can reduce temporal invariance, make temporal pooling implicitly coarse-position-aware, and expose multiscale interactions through channel mixing, while often improving computational efficiency by shortening th
Fast NF4 Dequantization Kernels for Large Language Model Inference
arXiv:2604.02556v1 Announce Type: new Abstract: Large language models (LLMs) have grown beyond the memory capacity of single GPU devices, necessitating quantization techniques for practical deployment. While NF4 (4-bit NormalFloat) quantization enables 4$\times$ memory reduction, inference on current NVIDIA GPUs (e.g., Ampere A100) requires expensive dequantization back to FP16 format, creating a critical performance bottleneck. This paper presents a lightweight shared memory optimization that addresses this gap through principled memory hierarchy exploitation while maintaining full ecosystem compatibility. We compare our technique against the open-source BitsAndBytes implementation, achieving 2.0--2.2$\times$ kernel speedup across three models (Gemma 27B, Qwen3 32B, and Llama3.3 70B) and
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!