# vector embeddings in AI

Vector embeddings are numerical representations that help machines process data, created by converting words, sentences, or images into vectors.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## What Is a Vector Database? How It Works and Why AI Needs One

DevFeed: [What Is a Vector Database? How It Works and Why AI Needs One](<https://devfeed.tech/articles/what-is-a-vector-database-how-it-works-and-why-ai-needs-one-145883.md>)

Original publisher: [Read original article](<https://montecarlo.ai/blog-what-is-a-vector-database>)

Author: Virna Sekuj

Published: 2026-11-19T17:18:16Z

Content type: article

Language: en

Sources: [Monte Carlo](<https://devfeed.tech/sources/monte-carlo.md>)

Topics: [Vectors & LLMs](<https://devfeed.tech/topics/vectors-llms.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>), [ann](<https://devfeed.tech/topics/ann.md>), [Database](<https://devfeed.tech/topics/database.md>), [information retrieval](<https://devfeed.tech/topics/information-retrieval.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>)

Tags: [agent-trust](<https://devfeed.tech/tags/agent-trust.md>), [agent-trust-glossary](<https://devfeed.tech/tags/agent-trust-glossary.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [data](<https://devfeed.tech/tags/data.md>), [database](<https://devfeed.tech/tags/database.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [glossary](<https://devfeed.tech/tags/glossary.md>), [rag](<https://devfeed.tech/tags/rag.md>), [rag-systems](<https://devfeed.tech/tags/rag-systems.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [retrieval-augmented-generation](<https://devfeed.tech/tags/retrieval-augmented-generation.md>), [retrieval-augmented-generation-rag](<https://devfeed.tech/tags/retrieval-augmented-generation-rag.md>), [vector](<https://devfeed.tech/tags/vector.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>), [vector-database-examples](<https://devfeed.tech/tags/vector-database-examples.md>), [vector-database-for-rag](<https://devfeed.tech/tags/vector-database-for-rag.md>), [vector-database-vs-relational-database](<https://devfeed.tech/tags/vector-database-vs-relational-database.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>), [vector-store](<https://devfeed.tech/tags/vector-store.md>)

### AI overview

Vector databases store content embeddings and use similarity search to retrieve relevant passages even when query wording differs. In a typical RAG flow, source content is chunked and embedded, stored with metadata, and indexed for approximate nearest-neighbor search; the user query is embedded and matching chunks are passed to the LLM as context. ANN indexing trades a small amount of precision for speed, and metadata filters can narrow results. Options include dedicated databases, database extensions, and vector search built into data platforms; choice depends on scale, latency, data location, and operational overhead. Retrieval can silently degrade when source syncs become stale, ingestion or chunking breaks, embedding models change without re-embedding, duplicate versions compete, or permission filters are missing. The text recommends monitoring freshness and ingestion volume, checking chunk quality, tracking embedding versions, measuring relevance and grounding, and tracing which chunks each agent request retrieves.

### Source excerpt

A vector database is a database built to store embeddings, which are numeric representations of the meaning of text, images, or other data, and to quickly find the embeddings most similar to a query. It's the retrieval layer behind most retrieval-augmented generation (RAG) systems and many AI agents, letting them find relevant information by meaning instead of exact keywords.

## FACET - Facet Aware Contrastive Embeddings for Traits- Big Five Estimation from Life Narratives

DevFeed: [FACET - Facet Aware Contrastive Embeddings for Traits- Big Five Estimation from Life Narratives](<https://devfeed.tech/articles/facet-facet-aware-contrastive-embeddings-for-traits-big-five-estimation-from-life-narratives-142670.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23560>)

Author: Balakrishnan Arumugam, Priyen Parekh, Mehak Gupta, Joshua Oltmanns

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Data Science](<https://devfeed.tech/topics/data-science.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [concatenation](<https://devfeed.tech/topics/concatenation.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [aging](<https://devfeed.tech/tags/aging.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [attention](<https://devfeed.tech/tags/attention.md>), [based](<https://devfeed.tech/tags/based.md>), [can](<https://devfeed.tech/tags/can.md>), [combines](<https://devfeed.tech/tags/combines.md>), [compare](<https://devfeed.tech/tags/compare.md>), [concatenation](<https://devfeed.tech/tags/concatenation.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [decision](<https://devfeed.tech/tags/decision.md>), [domain](<https://devfeed.tech/tags/domain.md>), [domains](<https://devfeed.tech/tags/domains.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [error](<https://devfeed.tech/tags/error.md>), [existing](<https://devfeed.tech/tags/existing.md>), [features](<https://devfeed.tech/tags/features.md>), [framework](<https://devfeed.tech/tags/framework.md>), [gives](<https://devfeed.tech/tags/gives.md>), [input](<https://devfeed.tech/tags/input.md>), [life](<https://devfeed.tech/tags/life.md>)

### AI overview

FACET estimates Big Five scores from life-narrative interviews by separately encoding text and psycholinguistic features, learning contrastive embeddings for personality facets, then using attention to pool them into domain predictions. In 1,332 SPAN participants, facet AUC reached 0.740 and domain R² reached 0.328, highest for Openness. End-to-end training had lower error than staged training in every domain, and attention pooling had the lowest error in every domain. Combining feature views reduced error versus text alone for Conscientiousness and Agreeableness; learned weighting showed no consistent advantage over concatenation. The method modestly outperformed raw-feature regression in four of five domains and provides facet-level predictions unavailable from a domain-only model.

### Source excerpt

arXiv:2611.23560v1 Announce Type: new Abstract: Personality is only weakly and inconsistently signaled by written language, and existing text-based methods typically combine language-model and psycholinguistic representations at the input or decision level. It remains open whether a more deliberate combination of the two views, and a decomposition of the prediction target, can recover more of the signal. We present a three-phase framework for estimating Big Five personality domain scores from life-narrative interview text. It encodes text and psycholinguistic features separately and combines them by concatenation or by a learned weighting, decomposes each domain into contrastively trained facet embeddings instead of predicting the domain directly, and pools the facet embeddings by attention into a domain-level prediction. We compare end-to-end with staged training and ablate the input modalities, their combination, and the facet-pooling strategy. On 1,332 participants from the St. Louis Personality and Aging Network (SPAN) study, facet embeddings reach AUC up to 0.740, and domain-level R2 reaches 0.328, highest for Openness. End-to-end training gives lower error than staged training in every domain, and attention pooling gives the lowest error in every domain. Combining the two views lowers error relative to text alone for Conscientiousness and Agreeableness, whereas a learned weighting of the views shows no consistent advantage over concatenation. The framework provides facet- level predictions that a domain-only model cannot, and it improves modestly on a regression from the raw features in four of the five domains.

## What Is an Embedding? How AI Turns Meaning Into Numbers

DevFeed: [What Is an Embedding? How AI Turns Meaning Into Numbers](<https://devfeed.tech/articles/what-is-an-embedding-how-ai-turns-meaning-into-numbers-141932.md>)

Original publisher: [Read original article](<https://montecarlo.ai/blog-embedding>)

Author: Ruslan Sultanov

Published: 2026-11-18T18:58:12Z

Content type: article

Language: en

Sources: [Monte Carlo](<https://devfeed.tech/sources/monte-carlo.md>)

Topics: [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [cosine similarity search](<https://devfeed.tech/topics/cosine-similarity-search.md>), [Data Quality](<https://devfeed.tech/topics/data-quality.md>), [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [clustering](<https://devfeed.tech/topics/clustering.md>)

Tags: [agent-trust](<https://devfeed.tech/tags/agent-trust.md>), [agent-trust-glossary](<https://devfeed.tech/tags/agent-trust-glossary.md>), [ai](<https://devfeed.tech/tags/ai.md>), [cosine-similarity](<https://devfeed.tech/tags/cosine-similarity.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embedding-model](<https://devfeed.tech/tags/embedding-model.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [embeddings-in-machine-learning](<https://devfeed.tech/tags/embeddings-in-machine-learning.md>), [embeddings-rag](<https://devfeed.tech/tags/embeddings-rag.md>), [glossary](<https://devfeed.tech/tags/glossary.md>), [meaning](<https://devfeed.tech/tags/meaning.md>), [text-embeddings](<https://devfeed.tech/tags/text-embeddings.md>), [vector-database](<https://devfeed.tech/tags/vector-database.md>), [vector-embeddings](<https://devfeed.tech/tags/vector-embeddings.md>), [what-is-an-embedding](<https://devfeed.tech/tags/what-is-an-embedding.md>)

### AI overview

Embeddings represent text, images, or other data as numeric vectors, placing semantically similar items near each other so systems can retrieve by meaning rather than exact wording. In a typical RAG pipeline, documents are split into chunks, embedded and indexed with metadata; a query is embedded to retrieve nearby passages for an LLM. Vectors are meaningful only within the model that produced them, and they become outdated when source content changes. Production reliability therefore depends on versioning vectors, re-embedding changed documents, removing deleted content, monitoring index freshness and evaluating and tracing retrieval. Stale or missing passages can lead an agent to give confident but outdated answers.

### Source excerpt

A customer types "can I send this back?" Your help center never uses that phrase. It talks about the "return window" and the "refund policy." Keyword search finds nothing. Yet a modern AI agent pulls up the right policy in milliseconds, because it isn't matching words. It's matching meaning. That's what embeddings do. This guide ...

## Comparative Evaluation of Embedding Representations for Financial News Sentiment Analysis

DevFeed: [Comparative Evaluation of Embedding Representations for Financial News Sentiment Analysis](<https://devfeed.tech/articles/comparative-evaluation-of-embedding-representations-for-financial-news-sentiment-analysis-137917.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2512.13749>)

Author: Joyjit Roy, Samaresh Kumar Singh

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Natural language processing](<https://devfeed.tech/topics/nlp.md>), [few-shot](<https://devfeed.tech/topics/few-shot.md>), [Finance](<https://devfeed.tech/topics/finance.md>)

Tags: [aggregation](<https://devfeed.tech/tags/aggregation.md>), [boosting](<https://devfeed.tech/tags/boosting.md>), [classification](<https://devfeed.tech/tags/classification.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ce](<https://devfeed.tech/tags/cs-ce.md>), [cs-cy](<https://devfeed.tech/tags/cs-cy.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [data-augmentation](<https://devfeed.tech/tags/data-augmentation.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [few-shot](<https://devfeed.tech/tags/few-shot.md>), [financial](<https://devfeed.tech/tags/financial.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [news](<https://devfeed.tech/tags/news.md>), [overfitting](<https://devfeed.tech/tags/overfitting.md>), [performance](<https://devfeed.tech/tags/performance.md>), [sentiment-analysis](<https://devfeed.tech/tags/sentiment-analysis.md>), [small](<https://devfeed.tech/tags/small.md>), [standard](<https://devfeed.tech/tags/standard.md>), [study](<https://devfeed.tech/tags/study.md>), [techniques](<https://devfeed.tech/tags/techniques.md>), [word2vec](<https://devfeed.tech/tags/word2vec.md>)

### AI overview

On 349 manually labeled financial news headlines, Word2Vec, GloVe, and sentence-transformer embeddings paired with gradient boosting showed a substantial gap between validation and test results and underperformed trivial baselines. The study attributes this in part to overfitting during model selection on small validation sets, and reports diminishing returns from pretrained embeddings when labeled data are scarce. It illustrates weekly sentiment aggregation and narrative summaries for market monitoring. The findings suggest embedding quality alone cannot overcome data scarcity; the abstract proposes few-shot learning, data augmentation, or lexicon-enhanced hybrids as alternatives to consider.

### Source excerpt

arXiv:2512.13749v3 Announce Type: replace-cross Abstract: Financial sentiment analysis enhances market understanding. However, standard Natural Language Processing (NLP) approaches encounter significant challenges when applied to small datasets. This study presents a comparative evaluation of embedding-based techniques for financial news sentiment classification in resource-constrained environments. Word2Vec, GloVe, and sentence transformer representations are evaluated in combination with gradient boosting on a manually labeled dataset of 349 financial news headlines. Experimental results identify a substantial gap between validation and test performance. Despite strong validation metrics, models underperform relative to trivial baselines. The analysis indicates that pretrained embeddings yield diminishing returns below a critical data sufficiency threshold. Small validation sets contribute to overfitting during model selection. Practical application is illustrated through weekly sentiment aggregation and narrative summarization for market monitoring. Overall, the findings indicate that embedding quality alone cannot address fundamental data scarcity in sentiment classification. Practitioners with limited labeled data should consider alternative strategies, including few-shot learning, data augmentation, or lexicon-enhanced hybrid methods.

## How IGNYTE runs candidate search with Amazon S3 Vectors

DevFeed: [How IGNYTE runs candidate search with Amazon S3 Vectors](<https://devfeed.tech/articles/how-ignyte-runs-candidate-search-with-amazon-s3-vectors-133435.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/storage/how-ignyte-runs-candidate-search-with-amazon-s3-vectors/>)

Author: Saurabh Gupta

Published: 2026-11-13T19:52:38Z

Content type: article

Language: en

Sources: [AWS Storage Blog](<https://devfeed.tech/sources/aws-storage-blog.md>)

Topics: [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [Amazon OpenSearch Service](<https://devfeed.tech/topics/amazon-opensearch-service.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Amazon Bedrock AgentCore](<https://devfeed.tech/topics/amazon-bedrock-agentcore.md>), [Amazon SageMaker](<https://devfeed.tech/topics/amazon-sagemaker.md>), [Parquet files in Amazon S3](<https://devfeed.tech/topics/parquet-files-in-amazon-s3.md>)

Tags: [agentcore](<https://devfeed.tech/tags/agentcore.md>), [amazon-opensearch-service](<https://devfeed.tech/tags/amazon-opensearch-service.md>), [amazon-sagemaker-ai](<https://devfeed.tech/tags/amazon-sagemaker-ai.md>), [amazon-simple-storage-service-s3](<https://devfeed.tech/tags/amazon-simple-storage-service-s3.md>), [approximate-nearest-neighbor-search](<https://devfeed.tech/tags/approximate-nearest-neighbor-search.md>), [aws-glue](<https://devfeed.tech/tags/aws-glue.md>), [aws-lambda](<https://devfeed.tech/tags/aws-lambda.md>), [customer-solutions](<https://devfeed.tech/tags/customer-solutions.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [intermediate-200](<https://devfeed.tech/tags/intermediate-200.md>), [layout](<https://devfeed.tech/tags/layout.md>), [metadata](<https://devfeed.tech/tags/metadata.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [semantic](<https://devfeed.tech/tags/semantic.md>)

### AI overview

IGNYTE moved embeddings for more than 1 billion contacts from OpenSearch Service to Amazon S3 Vectors; more than 800 million contacts are currently embedded. OpenSearch Service retains full profiles and handles structured filters. A scheduled batch pipeline loads company records first, then enriches contact records, generates embeddings, and writes profiles and vectors to their respective stores. Vectors are partitioned by continent, with minimal metadata for query filters. Searches embed the recruiter's query, retrieve the 100 closest contacts from the matching S3 Vectors index, then apply skills, experience, and other structured filters in OpenSearch Service. The article reports that the full agent matching workflow averages under 30 seconds after migration, compared with 3-4 minutes before; this is an end-to-end result, not a single vector query.

### Source excerpt

Ignyte moved the embeddings for more than 1 billion contacts from Amazon OpenSearch Service to Amazon S3 Vectors, keeping OpenSearch Service for structured filtering. See the per-continent index layout, the minimal-metadata design, and the batch pipeline behind its two-hop semantic candidate search.

## What Does Each Part of Speech Need? Contexts, Subspaces and Dimensions in Static Word Embeddings

DevFeed: [What Does Each Part of Speech Need? Contexts, Subspaces and Dimensions in Static Word Embeddings](<https://devfeed.tech/articles/what-does-each-part-of-speech-need-contexts-subspaces-and-dimensions-in-static-word-embeddings-132958.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.15114>)

Author: Jintang Xue, Chengwei Wei, Yun-Cheng Wang, C. -C. Jay Kuo

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [context](<https://devfeed.tech/tags/context.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [dimensions](<https://devfeed.tech/tags/dimensions.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [similarity](<https://devfeed.tech/tags/similarity.md>)

### AI overview

Across nine controlled corpora, part-of-speech tagging improves verb similarity on SimVerb-3500 at dimensions from 50 to 300, while noun gains are smaller. In verb-only text, adding adverbs helps verbs more than adding the same number of random words; adding adjectives gives nouns only slight gains. At a fixed dimension, separate PCA bases for each part of speech preserve verb and adjective similarity better than global or frequency-band bases. The study finds that nouns need more dimensions than verbs only among frequent words, and proposes allocating dimensions by frequency using these bases. At one-third of the original table size, only rare words lose dimensions; noun similarity on SimVerb-3500 and SimLex-999 matches the uncompressed table, and verb-to-noun retrieval remains within 0.02.

### Source excerpt

arXiv:2611.15114v1 Announce Type: cross Abstract: Static word embeddings give a word one vector, ignoring its different functions as noun, verb or adjective. Training on part-of-speech-tagged text gives each function its own vector, at the cost of a larger table. Using nine controlled corpora, we ask which context words each part of speech needs, whether it needs its own subspace, and whether the number of dimensions a word keeps should follow its part of speech or its frequency. The tags raise verb similarity on SimVerb-3500 at every dimension from 50 to 300 but help nouns much less. Added to verb-only text, adverbs help verbs more than as many random words; adjectives help nouns only slightly. At a fixed dimension, per-part-of-speech PCA bases preserve verb and adjective similarity better than a global basis or frequency-band bases. Nouns need more dimensions than verbs only among frequent words. We therefore allocate dimensions by frequency with these bases: at a third of the original size, only rare words lose dimensions, SimVerb-3500 and SimLex-999 nouns match the uncompressed table, and verb-to-noun retrieval stays within 0.02. Part of speech decides what a word's vector should be; frequency decides how much of it to keep.

## Sheaf-Inspired Structured Scoring Extensions of LightGCN for Collaborative Filtering

DevFeed: [Sheaf-Inspired Structured Scoring Extensions of LightGCN for Collaborative Filtering](<https://devfeed.tech/articles/sheaf-inspired-structured-scoring-extensions-of-lightgcn-for-collaborative-filtering-133193.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17226>)

Author: Ksenia Belkovskaya, Timur Lidzhiev, Maria Stebliy, Alexander Tarakanov

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>)

Tags: [amazon](<https://devfeed.tech/tags/amazon.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [catalog](<https://devfeed.tech/tags/catalog.md>), [cellular](<https://devfeed.tech/tags/cellular.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-ce](<https://devfeed.tech/tags/cs-ce.md>), [cs-cg](<https://devfeed.tech/tags/cs-cg.md>), [cs-ir](<https://devfeed.tech/tags/cs-ir.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [extensions](<https://devfeed.tech/tags/extensions.md>), [full](<https://devfeed.tech/tags/full.md>), [graph](<https://devfeed.tech/tags/graph.md>), [identity](<https://devfeed.tech/tags/identity.md>), [inner-product](<https://devfeed.tech/tags/inner-product.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [layer](<https://devfeed.tech/tags/layer.md>), [linear](<https://devfeed.tech/tags/linear.md>), [local](<https://devfeed.tech/tags/local.md>), [low-rank](<https://devfeed.tech/tags/low-rank.md>), [maps](<https://devfeed.tech/tags/maps.md>), [matrix](<https://devfeed.tech/tags/matrix.md>), [structured](<https://devfeed.tech/tags/structured.md>)

### AI overview

The study adds pair-dependent scoring transformations to a three-layer LightGCN recommender while leaving graph propagation unchanged. It compares identity, diagonal, rank-one low-rank, and full matrix residual scorers with a parameter-matched residual MLP on Amazon-Book, Gowalla, MovieLens-1M, and Yelp2018, measuring ranking quality and catalog coverage at cutoffs 10 and 20. Selected structured variants yield modest, dataset-dependent ranking changes and sometimes substantial coverage changes. The approach is sheaf-inspired but neither constructs a cellular sheaf nor uses sheaf-Laplacian diffusion, so the experiments do not establish a causal benefit from sheaf geometry.

### Source excerpt

arXiv:2611.17226v1 Announce Type: cross Abstract: Graph-based collaborative filtering learns user and item representations by propagating embeddings over a user--item interaction graph. We investigate a modular, sheaf-inspired scoring extension placed on top of a three-layer LightGCN backbone. The motivation comes from cellular sheaves, where local vector spaces are related by linear restriction maps; here, this idea motivates structured, pair-dependent transformations of user and item embeddings rather than a change to graph propagation. We compare the identity inner-product scorer, diagonal, rank-one low-rank, and full matrix residual scorers, and a parameter-matched residual MLP baseline (\texttt{mlp\_full}) on Amazon-Book, Gowalla, MovieLens-1M, and Yelp2018. We report ranking quality and catalog coverage at cutoffs 10 and 20. The selected structured variants show modest, dataset-dependent changes in ranking quality and sometimes substantial changes in coverage. Our method is inspired by sheaf geometry but does not construct a cellular sheaf or implement sheaf-Laplacian diffusion; the experiments therefore do not establish a causal benefit from sheaf geometry itself.

## How Hathr.AI delivers multi-tenant agentic RAG for regulated US government workloads with Amazon S3 Vectors

DevFeed: [How Hathr.AI delivers multi-tenant agentic RAG for regulated US government workloads with Amazon S3 Vectors](<https://devfeed.tech/articles/how-hathr-ai-delivers-multi-tenant-agentic-rag-for-regulated-us-government-workloads-with-amazon-s3-vectors-130617.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/storage/how-hathr-ai-delivers-multi-tenant-agentic-rag-for-regulated-us-government-workloads-with-amazon-s3-vectors/>)

Author: Sam Hart

Published: 2026-11-12T15:36:34Z

Content type: article

Language: en

Sources: [AWS Storage Blog](<https://devfeed.tech/sources/aws-storage-blog.md>)

Topics: [Multi-step AI processes](<https://devfeed.tech/topics/multi-step-ai-processes.md>), [tenant data protection](<https://devfeed.tech/topics/tenant-data-protection.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Amazon S3](<https://devfeed.tech/topics/amazon-s3.md>), [AWS Step Functions](<https://devfeed.tech/topics/aws-step-functions.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-rag](<https://devfeed.tech/tags/agentic-rag.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [amazon-documentdb](<https://devfeed.tech/tags/amazon-documentdb.md>), [amazon-s3](<https://devfeed.tech/tags/amazon-s3.md>), [amazon-simple-storage-service-s3](<https://devfeed.tech/tags/amazon-simple-storage-service-s3.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-govcloud-us](<https://devfeed.tech/tags/aws-govcloud-us.md>), [customer-solutions](<https://devfeed.tech/tags/customer-solutions.md>), [ecs](<https://devfeed.tech/tags/ecs.md>), [government](<https://devfeed.tech/tags/government.md>), [hipaa](<https://devfeed.tech/tags/hipaa.md>), [intermediate-200](<https://devfeed.tech/tags/intermediate-200.md>), [rag](<https://devfeed.tech/tags/rag.md>), [step-functions](<https://devfeed.tech/tags/step-functions.md>), [strands-agents](<https://devfeed.tech/tags/strands-agents.md>), [tool](<https://devfeed.tech/tags/tool.md>), [us-government](<https://devfeed.tech/tags/us-government.md>), [vector](<https://devfeed.tech/tags/vector.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>), [vectors](<https://devfeed.tech/tags/vectors.md>)

### AI overview

Hathr.AI moved high-churn vector search from Amazon DocumentDB to Amazon S3 Vectors, using a separate index and ARN for each customer in AWS GovCloud (US). Step Functions and Lambda parse, chunk, and embed uploads; Amazon DocumentDB retains text and metadata, while S3 Vectors stores embeddings. Deletion removes vectors immediately in batches before deleting chunks. IAM scopes ingestion and agent access to customer indexes, and authenticated session context selects the index. Strands Agents on ECS Fargate can search, inspect results, and retry with revised queries; Hathr.AI caps searches per turn. The move addressed shared-cluster resource contention and reduced the DocumentDB cluster to one quarter of peak size. Hathr.AI reports zero failed vector requests, sub-100 ms average query latency, and upload-to-search readiness P99 below one-tenth of its prior time. S3 Vectors has per-index write limits, and index settings such as dimensions, distance metric, and encryption cannot be changed after creation. GovCloud deployments also need partition-specific ARNs and endpoint checks, model availability confirmation, and appropriate data-event logging where audit records are needed.

### Source excerpt

Hathr.AI runs a HIPAA-aligned, multi-tenant agentic RAG tool for regulated industries in AWS GovCloud (US). This post covers its architecture (a Step Functions ingestion pipeline, per-customer Amazon S3 Vectors indexes, and Strands Agents on ECS Fargate) and the results after moving vector search from Amazon DocumentDB to Amazon S3 Vectors.

## Embeddings for Preferences, Not Semantics

DevFeed: [Embeddings for Preferences, Not Semantics](<https://devfeed.tech/articles/embeddings-for-preferences-not-semantics-124409.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2605.08360>)

Author: Carter Blair, Ariel D. Procaccia, Milind Tambe

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [decision-making](<https://devfeed.tech/topics/decision-making.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [clustering](<https://devfeed.tech/topics/clustering.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [clustering](<https://devfeed.tech/tags/clustering.md>), [collective](<https://devfeed.tech/tags/collective.md>), [correlation](<https://devfeed.tech/tags/correlation.md>), [cosine](<https://devfeed.tech/tags/cosine.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [fail](<https://devfeed.tech/tags/fail.md>), [invariance](<https://devfeed.tech/tags/invariance.md>), [literature](<https://devfeed.tech/tags/literature.md>), [location](<https://devfeed.tech/tags/location.md>), [models](<https://devfeed.tech/tags/models.md>), [modern](<https://devfeed.tech/tags/modern.md>), [online](<https://devfeed.tech/tags/online.md>), [opinions](<https://devfeed.tech/tags/opinions.md>), [semantics](<https://devfeed.tech/tags/semantics.md>)

### AI overview

Standard text embeddings capture semantic similarity, but facility-location and fair-clustering applications need preferential similarity: how much a participant agrees with a text, reflected as distance. Because stance and values can be correlated with wording and style, off-the-shelf embeddings may appear to capture preferences while relying on those semantic cues, and can fail when the correlation breaks. The paper frames this as an invariance problem and reports that synthetic training data designed to break the correlation provably shifts the optimal scorer away from nuisance-dominated cosine. It also reports significantly better preference prediction across 11 online deliberation datasets.

### Source excerpt

arXiv:2605.08360v2 Announce Type: replace Abstract: Modern AI is opening the door to collective decision-making in which participants express their views as free-form text rather than voting on a fixed set of candidates. A natural idea is to embed these opinions in a vector space so that the substantial literature on facility location problems and fair clustering can be brought to bear. But standard text embeddings measure semantic similarity, whereas distances in facility location problems and fair clustering require what we call preferential similarity: a participant's agreement with a piece of text should be inversely related to their distance from it. Off-the-shelf embeddings inherit a coarse preference signal through a correlation between semantic and preferential similarity, but fail to capture preferences when the correlation breaks. We formalize this as an invariance problem: text embedding models encode both a preference-relevant signal (stance and values) and semantic nuisance (style and wording), and the two are observationally correlated, so a geometry that relies on nuisance can appear preference-correct even when it is not. We show that synthetic training data designed to break this correlation provably shifts the optimal scorer away from nuisance-dominated cosine and significantly improves preference prediction across 11 online deliberation datasets.

## User Embeddings are Superpositions of Interpretable Behavioral Modes

DevFeed: [User Embeddings are Superpositions of Interpretable Behavioral Modes](<https://devfeed.tech/articles/user-embeddings-are-superpositions-of-interpretable-behavioral-modes-120757.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.08487>)

Author: Jonathan Kozaczuk, Zheng Dong, Noyan Evirgen, Keivan Majidi, Eddie K. H. Ng, Chris Porter, Qi Zhao, Jia Geng, Changmao Li, Brandon R. Feng

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [personalization](<https://devfeed.tech/topics/personalization.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [autoencoders](<https://devfeed.tech/tags/autoencoders.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [interpretability](<https://devfeed.tech/tags/interpretability.md>), [linear](<https://devfeed.tech/tags/linear.md>), [model](<https://devfeed.tech/tags/model.md>), [model-agnostic](<https://devfeed.tech/tags/model-agnostic.md>), [personalization](<https://devfeed.tech/tags/personalization.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [superposition](<https://devfeed.tech/tags/superposition.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [synthetic-data](<https://devfeed.tech/tags/synthetic-data.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [user](<https://devfeed.tech/tags/user.md>), [zero-shot](<https://devfeed.tech/tags/zero-shot.md>)

### AI overview

The paper proposes interpreting dense user embeddings as sparse superpositions of behavioral modes. It uses sparse dictionary learning, including sparse autoencoders (SAEs), to recover modes and assign natural-language labels. On synthetic data with known ground truth, SAEs recover modes better than classical sparse baselines, including for transformer-encoded embeddings. The authors also apply the method to classical and transformer-based embeddings from two real-world datasets, demonstrating zero-shot audience retrieval and characterization using labels, even though the embeddings were not trained on text. The approach is post-hoc and model-agnostic, so it requires no changes to the upstream embedding system.

### Source excerpt

arXiv:2611.08487v1 Announce Type: new Abstract: User embeddings are central components of modern recommendation, personalization, and behavioral modeling systems. Though these embeddings can compactly encode rich behavioral information, they are inherently dense and opaque, limiting interpretability. To address this gap, we argue that user embeddings can be understood as sparse superpositions of interpretable behavioral modes. This structure is directly predicted by generative models of user behavior for linear embeddings and extends approximately to deep-learning representations by the superposition hypothesis. Superimposed modes can naturally be recovered by sparse dictionary learning methods, such as sparse autoencoders (SAEs), and recovered modes can be automatically named and queried in natural language. We validate mode recovery against known ground truth on synthetic data, where SAEs recover modes better than classical sparse baselines, including on transformer-encoded embeddings. We then apply the method to classical and transformer-based user embeddings from two real-world datasets. We also show how the corresponding natural-language labels enable zero-shot audience retrieval and characterization over embeddings never trained on text. The framework is post-hoc and model-agnostic, requiring no modification to the upstream embedding system.

## BrainBind: Bridging Synchronized EEG and fMRI through Long-Form Naturalistic Video Representations

DevFeed: [BrainBind: Bridging Synchronized EEG and fMRI through Long-Form Naturalistic Video Representations](<https://devfeed.tech/articles/brainbind-bridging-synchronized-eeg-and-fmri-through-long-form-naturalistic-video-representations-118348.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.05581>)

Author: Shixiao Wang, Rining Wu, Jian K. Liu

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [multimodal-ai](<https://devfeed.tech/topics/multimodal-ai.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [context](<https://devfeed.tech/tags/context.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [eeg](<https://devfeed.tech/tags/eeg.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [framework](<https://devfeed.tech/tags/framework.md>), [neural](<https://devfeed.tech/tags/neural.md>), [recall](<https://devfeed.tech/tags/recall.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [spaces](<https://devfeed.tech/tags/spaces.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [static](<https://devfeed.tech/tags/static.md>), [stimulus](<https://devfeed.tech/tags/stimulus.md>), [system](<https://devfeed.tech/tags/system.md>), [video](<https://devfeed.tech/tags/video.md>), [vision](<https://devfeed.tech/tags/vision.md>)

### AI overview

BrainBind maps synchronized EEG and fMRI recorded during long-form naturalistic video viewing into frozen video-embedding spaces, using them as shared visual-semantic references rather than literal neural codes. On the CineBrain dataset, a five-seed comparison found different retrieval strengths across configurations: with Omni, fMRI most strongly organized the full candidate list (MRR 23.82; nDCG@10 25.31), while Joint-Dual achieved the highest mean label-aware R@1 (13.52%). An episode-disjoint evaluation examined how repeated stimulus identity and narrative context contribute to performance on a random split. The reported results show distinct top-rank and rank-wide retrieval profiles for the tested proxies and configurations in this dataset.

### Source excerpt

arXiv:2611.05581v1 Announce Type: new Abstract: Real-world vision is dynamic and scene-rich, while EEG and fMRI resolve it at complementary temporal and spatial scales. Most neural decoding systems, however, align one modality in isolation and mostly against static images or short clips. We study the synchronized case: EEG and fMRI are recorded during the same long-form naturalistic viewing, giving a shared stimulus but also recurring narrative context and many semantically similar candidates. We present BrainBind, a framework that maps synchronized EEG and fMRI into frozen long-form video embedding spaces and treats those spaces as common visual-semantic references, not literal neural codes. On CineBrain, a five-seed Recall@K 2 x 4 matrix across two video anchors and four system configurations shows distinct retrieval profiles. With Omni, fMRI organizes the full candidate list most strongly (MRR = 23.82, nDCG@10 = 25.31), while Joint-Dual gives the highest mean label-aware R@1 (13.52%) with a distinct rank-wide profile. The complementary episode-disjoint evaluation quantifies the contribution of repeated stimulus identity and narrative context to the random split. Across these experiments, proxy choice and system configuration produce distinct top-rank and rank-wide retrieval profiles within the present synchronized dataset. The code is available at https://github.com/Forainest789/Brainbind.

## Multimodal Fusion Methods for Whole-Slide Images and Clinical Data in Inflammatory Bowel Disease

DevFeed: [Multimodal Fusion Methods for Whole-Slide Images and Clinical Data in Inflammatory Bowel Disease](<https://devfeed.tech/articles/multimodal-fusion-methods-for-whole-slide-images-and-clinical-data-in-inflammatory-bowel-disease-118347.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.05579>)

Author: Ankana Banerjee (University of Stavanger, Norway), Mark van der Giezen (University of Stavanger, Norway), Ole Gunnar Aasprong (Stavanger University Hospital, Norway), Emiel A. M. Janssen (University o

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Medical imaging](<https://devfeed.tech/topics/medical-imaging.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [attention](<https://devfeed.tech/tags/attention.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [disease](<https://devfeed.tech/tags/disease.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [fusion](<https://devfeed.tech/tags/fusion.md>), [images](<https://devfeed.tech/tags/images.md>), [modular](<https://devfeed.tech/tags/modular.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [prediction](<https://devfeed.tech/tags/prediction.md>)

### AI overview

The study compares 15 fusion methods for combining whole-slide image and per-visit clinical embeddings to predict STRIDE-II-aligned remission outcomes in a prospective cohort of newly diagnosed inflammatory bowel disease patients. Slide embeddings come from an attention-based aggregator over patch embeddings from a frozen foundation model; the methods share an output dimension and prediction head except for decision-level methods. The best method achieved macro-F1 scores of 0.57-0.75 and outperformed both single-modality models in four of six scenarios, with the top method varying by outcome and input visit. Removing slide embeddings generally reduced balanced accuracy most for histological remission and least for endoscopic remission; methods with similar accuracy showed different degrees of reliance on image data.

### Source excerpt

arXiv:2611.05579v1 Announce Type: cross Abstract: Multimodal AI fusion of histology and clinical data is an active research area in computational pathology, but fusion methods are rarely compared under identical conditions and across different clinical endpoints. We present a modular framework in which slide-level embeddings, obtained by training an attention-based aggregator over patch embeddings from a frozen foundation model, and per-visit clinical embeddings are fed into 15 interchangeable fusion methods with a shared output dimension and prediction head (except decision-level methods). Using a prospective cohort of newly diagnosed inflammatory bowel disease patients, we predicted STRIDE-II-aligned remission outcomes from baseline or follow-up visits. The best fusion method reached macro-F1 of 0.57-0.75, outperforming both single-modality models in four of six scenarios, and differed by outcome and input. Image reliance, the drop in balanced accuracy when the slide embedding is removed, tended to follow the remission type - highest for histological and lowest for endoscopic remission; methods with similar accuracy relied on the slide to very different degrees.

## Adaptive Land Units from Earth Embeddings

DevFeed: [Adaptive Land Units from Earth Embeddings](<https://devfeed.tech/articles/adaptive-land-units-from-earth-embeddings-114484.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.01050>)

Author: Albert Alonso, Jamal Jokar Arsanjani, Andres R. Masegosa

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [geospatial](<https://devfeed.tech/topics/geospatial.md>), [clustering](<https://devfeed.tech/topics/clustering.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>)

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [bayesian](<https://devfeed.tech/tags/bayesian.md>), [case-study](<https://devfeed.tech/tags/case-study.md>), [clustering](<https://devfeed.tech/tags/clustering.md>), [consumer](<https://devfeed.tech/tags/consumer.md>), [cost](<https://devfeed.tech/tags/cost.md>), [country](<https://devfeed.tech/tags/country.md>), [crop](<https://devfeed.tech/tags/crop.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [denmark](<https://devfeed.tech/tags/denmark.md>), [device](<https://devfeed.tech/tags/device.md>), [earth](<https://devfeed.tech/tags/earth.md>), [eess-iv](<https://devfeed.tech/tags/eess-iv.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [fewer](<https://devfeed.tech/tags/fewer.md>), [field](<https://devfeed.tech/tags/field.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [grid](<https://devfeed.tech/tags/grid.md>), [half](<https://devfeed.tech/tags/half.md>), [images](<https://devfeed.tech/tags/images.md>), [large](<https://devfeed.tech/tags/large.md>), [units](<https://devfeed.tech/tags/units.md>)

### AI overview

The study builds adaptive land units by clustering AlphaEarth embeddings with parallel Bayesian hierarchical clustering: units are larger over uniform land and smaller where land is heterogeneous. In Denmark, this unsupervised method reduced 402 million pixels to 1.27 million units, enabling whole-country processing on one consumer-grade GPU. The authors report that regular grids needed about five times as many cells to match land-use labels, and about four times as many to match declared field crops; clustering embeddings achieved the same purity with half as many units as clustering satellite images directly. Land-use maps on these units were within a few accuracy points of pixel-based models at much lower cost. Whole-country estimates with reliable uncertainty took under a minute, while the pixel model did not fit in device memory. These results are reported for Denmark in the supplied abstract.

### Source excerpt

arXiv:2611.01050v1 Announce Type: cross Abstract: Modeling land use starts with a choice of spatial units, typically a regular grid of pixels whose resolution is set by the spatial detail the process requires and, often more critically, by the computational budget. With Earth observation data more abundant than ever, we ask whether good land units can instead emerge from the data itself. To do that, we cluster AlphaEarth embeddings with a parallel Bayesian hierarchical clustering into adaptive land units, large where the land is uniform and small where it is highly heterogeneous. Using Denmark as a case study, we turn 402M land pixels into 1.27M newly defined units, about 320 times fewer, without any labels (i.e., unsupervised), making it possible to fit the whole country on a single consumer-grade GPU. Importantly, these units describe the land well, as a regular grid needs about 5 times as many cells to match them against land-use labels, and about 4 times as many against the declared crop of every field. Compared to clustering the satellite images themselves, the embeddings need half the units for the same purity. But more notably, they work as modeling units. Land-use maps learned on them are within a few points of accuracy of pixel models at a small fraction of the cost, and they make whole-country land-use estimates with reliable uncertainty possible in under a minute, where the same model on pixels does not even fit in device memory.

## Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations

DevFeed: [Gluing Local Contexts into Global Meaning: A Sheaf-Theoretic Decomposition of Transformer Representations](<https://devfeed.tech/articles/gluing-local-contexts-into-global-meaning-a-sheaf-theoretic-decomposition-of-transformer-representations-111456.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.17117>)

Author: Bryce Grant, Peng Wang

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [cellular](<https://devfeed.tech/tags/cellular.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cycles](<https://devfeed.tech/tags/cycles.md>), [graph](<https://devfeed.tech/tags/graph.md>), [interpretability](<https://devfeed.tech/tags/interpretability.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [layers](<https://devfeed.tech/tags/layers.md>), [models](<https://devfeed.tech/tags/models.md>), [steering](<https://devfeed.tech/tags/steering.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

The paper uses paraphrase pairs as local consistency witnesses to construct a cellular sheaf over a paraphrase graph, then analyzes its Laplacian without training the frozen model. Its kernel defines a content subspace that, at 20 dimensions, exceeds the full-dimensional LEACE-retained subspace on held-out CounterFact retrieval by an average 17.1 percentage points across eight models; jointly constraining a fact's paraphrases adds 6.5 points. The spectral complement identifies coordinates linked to causal influence: ablating the most paraphrase-variable coordinates changes outputs 5.6-26.5 times more than ablating coordinates matched for per-coordinate variability, across nine models from 124M to 13B parameters. A layer-stacked grid yields harmonic mass that varies by nearly three orders of magnitude across six architectures and correlates with steering fragility (Spearman's ρ=0.90, one-sided permutation p=0.014). These are abstract-reported results; the supplied text does not include further methodological detail.

### Source excerpt

arXiv:2610.17117v2 Announce Type: replace-cross Abstract: Prior research into interpretability does not test whether a representation stays consistent under paraphrasing. Each pair of paraphrases can be thought of as a local witness that two activations should be the same. From these local witnesses we can build a cellular sheaf on the paraphrase graph. The question of what is consistent under paraphrasing can be answered by the Laplacian $L_\mathcal{F}$ given a frozen model. No training required. Our kernel $H^0$ gives us a content subspace. We show that with only 20 dims the kernel outperforms the subspace that LEACE retains with all dims on held-out CounterFact fact retrieval by $+17.1$\,pp on average across 8 models. By constraining all paraphrases of a fact together we see an additional increase of $+6.5$\,pp. Our spectral complement shows us where the causal influence is concentrated. Ablating the coordinates that vary the most when paraphrased (which approximates ablating the spectral complement at the coordinate level) leads to $5.6$ to $26.5\times$ larger changes in model outputs compared to ablating coordinates with matched per-coordinate variability. We find this across 9 models from 124M to 13B params. By stacking layers we can create a grid which has $b_1 = N(L-1)$ algebraic 1-cycles. The harmonic mass derived from the model varies by nearly 3 orders of magnitude between the 6 architectures. It also tracks the order of steering fragility (Spearman's $\rho = 0.90$, exact one-sided permutation $p = 0.014$). The kernel, its spectral complement and the harmonic mass all come from the same paraphrase graph. Project page: https://cwru-aism.github.io/gluing-lc-page/.

## Task Vectors Can Be Linearly Decomposed into Interpretable and Controllable Components

DevFeed: [Task Vectors Can Be Linearly Decomposed into Interpretable and Controllable Components](<https://devfeed.tech/articles/task-vectors-can-be-linearly-decomposed-into-interpretable-and-controllable-components-111621.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.41743>)

Author: Yukiya Horiba, Hakaze Cho, Takahiro Yakoh

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Hyperparameter optimization](<https://devfeed.tech/topics/hyperparameter-optimization.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [bayesian](<https://devfeed.tech/tags/bayesian.md>), [classification](<https://devfeed.tech/tags/classification.md>), [components](<https://devfeed.tech/tags/components.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [functional](<https://devfeed.tech/tags/functional.md>), [hidden-states](<https://devfeed.tech/tags/hidden-states.md>), [language](<https://devfeed.tech/tags/language.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models](<https://devfeed.tech/tags/large-language-models.md>), [learning](<https://devfeed.tech/tags/learning.md>), [models](<https://devfeed.tech/tags/models.md>), [multiple](<https://devfeed.tech/tags/multiple.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [optimize](<https://devfeed.tech/tags/optimize.md>), [original](<https://devfeed.tech/tags/original.md>), [performance](<https://devfeed.tech/tags/performance.md>), [random](<https://devfeed.tech/tags/random.md>), [recognition](<https://devfeed.tech/tags/recognition.md>), [search](<https://devfeed.tech/tags/search.md>), [steering](<https://devfeed.tech/tags/steering.md>)

### AI overview

The study decomposes a task vector (TV) into task-recognition (TR), task-learning (TL), and residual components using candidate-label unembeddings, then uses Bayesian optimization to tune TR and TL coefficients. Across multiple model families, classification tasks, and instruction- and demonstration-based TVs, optimized TR/TL directions outperform the original TVs. Random-direction optimization can also improve performance, but TR/TL search produces more reliable gains, especially for instruction-based TVs. Separate interventions on the components yield distinct logit-level effects consistent with their proposed roles. The results suggest TVs can support more interpretable and controllable interventions, though the evidence provided is an abstract rather than full paper text.

### Source excerpt

arXiv:2610.41743v1 Announce Type: new Abstract: Task Vectors (TVs) provide a compact way to steer large language models, but they are typically treated as atomized directions merely extracted from hidden states, leaving their internal functional structure poorly understood. Motivated by prior work that decomposes in-context learning into Task Recognition (TR) and Task Learning (TL), we ask whether a single TV can likewise be decomposed into functionally meaningful components. We formulate a TR/TL-based decomposition from candidate-label unembeddings, separating each TV into TR, TL, and residual components, and try to optimize the TR and TL coefficients using Bayesian optimization for more effective intervention. We evaluate this formulation across multiple model families, classification tasks, and both instruction- and demonstration-based TVs. Across these settings, TR/TL-based optimization identifies intervention directions that outperform the original TVs. Although optimization over random directions can also improve performance, TR/TL-based search yields more reliable gains, particularly for instruction-based TVs. Moreover, isolated interventions on the decomposed components produce distinct logit-level effects that are consistent with their intended functional roles. Together, these results suggest that TVs are not monolithic steering directions, but can be decomposed into functionally meaningful components that support both interpretation and control.

## How to Build a LlamaIndex Vector Store With TiDB

DevFeed: [How to Build a LlamaIndex Vector Store With TiDB](<https://devfeed.tech/articles/how-to-build-a-llamaindex-vector-store-with-tidb-109548.md>)

Original publisher: [Read original article](<https://www.pingcap.com/blog/llamaindex-vector-store/>)

Author: Akshata Hire

Published: 2026-10-30T19:16:02Z

Content type: tutorial

Language: en

Sources: [TiDB](<https://devfeed.tech/sources/tidb.md>)

Topics: [llamaindex](<https://devfeed.tech/topics/llamaindex.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [database-optimization](<https://devfeed.tech/topics/database-optimization.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>)

Tags: [build](<https://devfeed.tech/tags/build.md>), [can](<https://devfeed.tech/tags/can.md>), [content](<https://devfeed.tech/tags/content.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [first](<https://devfeed.tech/tags/first.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [index](<https://devfeed.tech/tags/index.md>), [llamaindex](<https://devfeed.tech/tags/llamaindex.md>), [memory](<https://devfeed.tech/tags/memory.md>), [metadata](<https://devfeed.tech/tags/metadata.md>), [node](<https://devfeed.tech/tags/node.md>), [node-embeddings](<https://devfeed.tech/tags/node-embeddings.md>), [production](<https://devfeed.tech/tags/production.md>), [queries](<https://devfeed.tech/tags/queries.md>), [store](<https://devfeed.tech/tags/store.md>), [systems](<https://devfeed.tech/tags/systems.md>), [tidb](<https://devfeed.tech/tags/tidb.md>), [tutorial](<https://devfeed.tech/tags/tutorial.md>), [vector](<https://devfeed.tech/tags/vector.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

TiDBVectorStore gives LlamaIndex a persistent, shared SQL-backed store for node embeddings, text, and JSON metadata; the existing VectorStoreIndex flow stays the same by passing it through StorageContext. The tutorial covers loading and chunking documents, matching vector_dimension to the embedding model, indexing and reconnecting, and filtering queries by fields such as tenant or document type. TiDB metadata filters run before distance ordering, returning exact filtered results but scanning matching rows, so cost grows with their number. TiDBVectorStore supports LlamaIndex's default query mode only; keyword or hybrid retrieval needs a separate path. TiDB full-text search uses BM25 but is limited to TiDB Cloud Starter in five AWS regions and does not guarantee exact phrase matching. The text also notes vector-search version requirements and public-preview status, and recommends evaluating top-k retrieval, planning tenant isolation, and tuning inserts for packet limits.

### Source excerpt

A LlamaIndex vector store is where your index keeps node embeddings, usually alongside the text chunks and metadata, so queries can find similar content. The default store lives in memory, so a restart or redeploy wipes the index and sends you back to your embedding provider to pay for the same corpus again. Production systems [...] The post How to Build a LlamaIndex Vector Store With TiDB appeared first on TiDB.

## PALATE: Pretrained Autoregressive LATent Embeddings

DevFeed: [PALATE: Pretrained Autoregressive LATent Embeddings](<https://devfeed.tech/articles/palate-pretrained-autoregressive-latent-embeddings-108767.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.39090>)

Author: Sizhang Zhao, Praveen Gondhi, Bin Li, Utsaw Kumar

Published: 2026-10-30T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [search-engines](<https://devfeed.tech/topics/search-engines.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>)

Tags: [ads](<https://devfeed.tech/tags/ads.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [autoregressive](<https://devfeed.tech/tags/autoregressive.md>), [business](<https://devfeed.tech/tags/business.md>), [commerce](<https://devfeed.tech/tags/commerce.md>), [consumer](<https://devfeed.tech/tags/consumer.md>), [conversion](<https://devfeed.tech/tags/conversion.md>), [counts](<https://devfeed.tech/tags/counts.md>), [cs-ir](<https://devfeed.tech/tags/cs-ir.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [daily](<https://devfeed.tech/tags/daily.md>), [delivery](<https://devfeed.tech/tags/delivery.md>), [design](<https://devfeed.tech/tags/design.md>), [discovery](<https://devfeed.tech/tags/discovery.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [engagement](<https://devfeed.tech/tags/engagement.md>), [eval](<https://devfeed.tech/tags/eval.md>), [exploration](<https://devfeed.tech/tags/exploration.md>), [feature](<https://devfeed.tech/tags/feature.md>), [food](<https://devfeed.tech/tags/food.md>), [generative](<https://devfeed.tech/tags/generative.md>)

### AI overview

PALATE creates a shared consumer embedding by training a GPT-style model to predict future engagements and using its final-token hidden state. A daily job refreshes this frozen vector, which rankers on browse and search surfaces consume as a dense feature. To adapt training to repurchase-heavy marketplaces, PALATE weights engagement with new businesses more heavily, pretrains on clicks before fine-tuning on conversions, and combines popular-business and shown-but-not-ordered negatives with random and in-batch negatives. Offline relative Eval-AUC gains were +0.59% for browse and +0.57% for search. Online, the embedding was one component of bundled ranker retrains; Discovery reported revenue +0.79% and conversion rate +1.17%, while Search reported revenue +0.94% and conversion rate +1.50%.

### Source excerpt

arXiv:2610.39090v1 Announce Type: cross Abstract: Local-commerce and food-delivery marketplaces run large ads businesses: across hundreds of thousands of stores, merchants pay to be surfaced on browse and search surfaces, and a ranker decides which ads each consumer sees. Ranking well demands strong per-consumer personalization, yet a consumer's raw interaction history is sparse and high-cardinality. Our production ranker had personalized only through hand-engineered signals: aggregate engagement counts, categorical embeddings, and order-agnostic pooled history, with no learned per-consumer sequence representation. We learn one by training a GPT-style generative model of the consumer's engagement sequence to predict their next few engagements, and read its last-token hidden state as the consumer embedding. That single frozen vector becomes one shared personalization input for every ad surface: a daily job refreshes it, and each ranker consumes it as an ordinary dense feature. Offline it is the largest single-feature gain in our ranking-AUC ladder, a relative Eval-AUC gain of +0.59% on a browse-driven feed and +0.57% on a query-driven search surface. Online, it shipped as one component of bundled ranker retrains on both surfaces: Discovery (revenue +0.79%, conversion rate +1.17%) and Search (revenue +0.94%, conversion rate +1.50%). Three design choices adapt the generative objective to a repurchase-heavy marketplace, and are the core of our contribution: an exploration-weighted loss that up-weights businesses new to the consumer over ones already engaged; a two-stage curriculum: a click pretrain that keeps the representation broad and exploratory, then a conversion fine-tune that sharpens purchase intent; and a designed negative set that pairs a fixed popular-business pool and businesses shown but not ordered with random and in-batch negatives.

## Cross-modal Steering Vectors Exist Across Both Image and Text Generation

DevFeed: [Cross-modal Steering Vectors Exist Across Both Image and Text Generation](<https://devfeed.tech/articles/cross-modal-steering-vectors-exist-across-both-image-and-text-generation-103933.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.36269>)

Author: Donghao Zeng (Mila, McGill University), Nishant Subramani (Carnegie Mellon University), Tamar Rott Shaham (MIT CSAIL), Benno Krojer (Mila, McGill University, Northeastern University)

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [multimodal-ai](<https://devfeed.tech/topics/multimodal-ai.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [alignment](<https://devfeed.tech/tags/alignment.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [concepts](<https://devfeed.tech/tags/concepts.md>), [cross](<https://devfeed.tech/tags/cross.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [design](<https://devfeed.tech/tags/design.md>), [effect](<https://devfeed.tech/tags/effect.md>), [emotion](<https://devfeed.tech/tags/emotion.md>), [formal](<https://devfeed.tech/tags/formal.md>), [gap](<https://devfeed.tech/tags/gap.md>), [generation](<https://devfeed.tech/tags/generation.md>), [image](<https://devfeed.tech/tags/image.md>), [image-generation](<https://devfeed.tech/tags/image-generation.md>), [language](<https://devfeed.tech/tags/language.md>), [models](<https://devfeed.tech/tags/models.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [multimodal-models](<https://devfeed.tech/tags/multimodal-models.md>), [object](<https://devfeed.tech/tags/object.md>), [steering](<https://devfeed.tech/tags/steering.md>), [text](<https://devfeed.tech/tags/text.md>)

### AI overview

The study finds that steering vectors can control image and text generation across modalities in unified multimodal models. Vectors learned to steer one modality transfer zero-shot in both directions, including for emotion and object size. For a modality-specific concept such as formal versus colloquial language, steering can produce an interpretable counterpart in the other modality, such as formal clothing or design. Although text and image representations have a clear modality gap, vectors with stronger cross-modal transfer tend to point in similar directions for the same concept. The evidence is limited to the abstract and does not specify model names or experimental details.

### Source excerpt

arXiv:2610.36269v1 Announce Type: cross Abstract: Are concepts and mechanisms unified across modalities in current AI models, or instead implemented separately for each one? We operationalize this broad question of modality unification by studying cross-modal steering: 1) finding vectors that successfully steer generation on tokens of one modality, which then 2) transfer zero-shot to steer generation when added to tokens of another modality. Our study goes beyond simple correlational alignment (e.g. representational similarity) because steering is fundamentally causal. We study cross-modal steering in several unified multimodal models (UMMs), since they possess shared parameters for language and image generation, focusing on these two modalities. Across diverse steered concepts such as emotion and object size, we show that steering vectors transfer zero-shot in both directions (text-to-image & image-to-text). Even for concepts that are highly unique to one modality (e.g. formal vs. colloquial language), steering vectors often induce a consistent, interpretable effect in the other modality, such as formal clothing and design. This transfer has a geometric signature too: despite a clear modality gap separating text and image representations, steering vectors with strong cross-modal transferability tend to point in similar directions for the same concept. Our findings shed light on how vision and language interact inside unified multimodal models and point towards more general, modality-agnostic approaches for controlling their behavior.

## Beyond the Readout: A Latent Numerical Frame Shared Across Modality and Operation

DevFeed: [Beyond the Readout: A Latent Numerical Frame Shared Across Modality and Operation](<https://devfeed.tech/articles/beyond-the-readout-a-latent-numerical-frame-shared-across-modality-and-operation-103268.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37422>)

Author: Oriol Juan Sabater, Alex Ferrando de las Morenas, Pau Rodr\\'iguez, Xavier Suau

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [addition](<https://devfeed.tech/tags/addition.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [gemma4](<https://devfeed.tech/tags/gemma4.md>), [hidden](<https://devfeed.tech/tags/hidden.md>), [image](<https://devfeed.tech/tags/image.md>), [metrics](<https://devfeed.tech/tags/metrics.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [number](<https://devfeed.tech/tags/number.md>), [numbers](<https://devfeed.tech/tags/numbers.md>), [rotation](<https://devfeed.tech/tags/rotation.md>), [stream](<https://devfeed.tech/tags/stream.md>), [text](<https://devfeed.tech/tags/text.md>), [time](<https://devfeed.tech/tags/time.md>), [upstream](<https://devfeed.tech/tags/upstream.md>), [write](<https://devfeed.tech/tags/write.md>)

### AI overview

Causal interventions in Gemma4-12B-it suggest that numerical representations for text and image inputs share a common geometry in different coordinate frames. Reusing a text-learned subspace on images recovered 53.2% of its causal effect; rotation and rescaling raised recovery to 91.4%. Within each subspace, 13 directions hidden from the output layer still carried number information while barely affecting the digit currently emitted. Swapping those directions redirected the next digit in 77.6% of cases when the numbers shared their first digit. The findings support a model that represents a multi-digit result before emitting it one digit at a time. The evidence here is the supplied abstract.

### Source excerpt

arXiv:2610.37422v1 Announce Type: new Abstract: When a multimodal model computes 17+35 from text or 91-39 from a rendered image, it produces the same answer, 52. Does it use one internal concept of number for both, or separate machinery for each path? Similarity metrics such as RSA or CKA cannot settle this question. They give the same score to identical and to rotated directions, and two representations can look alike only because both prepare the same output digits. We therefore use causal interventions. For four conditions (text or image, addition or subtraction), we learn a small subspace of Gemma4-12B-it's residual stream in which swapping coordinates swaps the model's answer. The subspaces answer three questions. First, do the conditions write numbers along the same directions? They do not. Reusing a subspace from text in an image on the same operation recovers only 53.2% of its causal effect. A single rotation and rescaling maps one subspace onto the other and raises recovery to 91.4%, so the conditions share one geometry expressed in different coordinates. Second, does this shared geometry only reflect the digit the model is about to print? To test this, we split each subspace into 9 directions that the output layer reads to choose the next digit and 13 directions that it cannot see. The 13 hidden directions still carry the number, yet changing them barely alters the digit being written. Third, what are these hidden directions for? A model must emit a multi-digit answer one digit at a time, so it can compute the full result early and keep it in reserve. When two numbers share their first digit, swapping the number in the hidden directions leaves the digit being written intact and redirects the next digit in 77.6% of cases. The model thus forms an abstract numerical concept upstream and unfolds it across time: multimodal representations share meaning through relational geometry across distinct coordinate frames.

## Tracing Contextual Token Representations Through Transformer Layers

DevFeed: [Tracing Contextual Token Representations Through Transformer Layers](<https://devfeed.tech/articles/tracing-contextual-token-representations-through-transformer-layers-100885.md>)

Original publisher: [Read original article](<https://hackernoon.com/tracing-contextual-token-representations-through-transformer-layers?source=rss>)

Author: Lian Kim

Published: 2026-10-26T17:22:30Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [Language models](<https://devfeed.tech/topics/language-models.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [model architecture](<https://devfeed.tech/topics/model-architecture.md>), [mlp](<https://devfeed.tech/topics/mlp.md>), [bert](<https://devfeed.tech/topics/bert.md>), [Data visualization](<https://devfeed.tech/topics/data-visualization.md>), [Temperature settings in AI](<https://devfeed.tech/topics/temperature-settings-in-ai.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [2019](<https://devfeed.tech/tags/2019.md>), [2020](<https://devfeed.tech/tags/2020.md>), [2023](<https://devfeed.tech/tags/2023.md>), [2025](<https://devfeed.tech/tags/2025.md>), [22](<https://devfeed.tech/tags/22.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [absolute](<https://devfeed.tech/tags/absolute.md>), [accounting](<https://devfeed.tech/tags/accounting.md>), [activation](<https://devfeed.tech/tags/activation.md>), [ai](<https://devfeed.tech/tags/ai.md>), [classification](<https://devfeed.tech/tags/classification.md>), [contextual-embeddings](<https://devfeed.tech/tags/contextual-embeddings.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [hackernoon-top-story](<https://devfeed.tech/tags/hackernoon-top-story.md>), [hidden](<https://devfeed.tech/tags/hidden.md>), [hidden-states](<https://devfeed.tech/tags/hidden-states.md>), [intent-classification](<https://devfeed.tech/tags/intent-classification.md>), [layers](<https://devfeed.tech/tags/layers.md>), [pythia-160m](<https://devfeed.tech/tags/pythia-160m.md>), [structured](<https://devfeed.tech/tags/structured.md>), [token](<https://devfeed.tech/tags/token.md>), [token-embeddings](<https://devfeed.tech/tags/token-embeddings.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [transformer-models](<https://devfeed.tech/tags/transformer-models.md>)

### AI overview

In a six-context Pythia-160M experiment, the same token `Ġlight` starts from an identical embedding, then shows a within-group distinction by recorded state 3 that persists through intermediate layers. Direction-based separation shrinks at the final state, while raw Euclidean separation grows alongside vector norms; the six selected contexts do not isolate word sense or establish semantic understanding. A separate banking-intent comparison shows interface-level trade-offs: on 320 eight-category messages, JEV had the highest strict accuracy and full label validity but the longest mean client latency; Claude was fastest but produced 22 invalid labels, and a post hoc normalization improved its score without surpassing JEV. The systems use different underlying models, runs, and billing units, so results do not isolate interface effects or establish direct cost or production-latency comparisons. Together, the experiments show why representation metrics and application outputs should be evaluated against the specific task and output contract.

### Source excerpt

Explore two AI experiments investigating contextual token representations, Transformer hidden states, and the accuracy of structured classification outputs.

## ChromaDB: Embeddings and Vector Databases in Python

DevFeed: [ChromaDB: Embeddings and Vector Databases in Python](<https://devfeed.tech/articles/chromadb-embeddings-and-vector-databases-in-python-100528.md>)

Original publisher: [Read original article](<https://realpython.com/chromadb-vector-database/>)

Author: Harrison Hoffman

Published: 2026-10-26T14:00:00Z

Content type: tutorial

Language: en

Sources: [Real Python](<https://devfeed.tech/sources/real-python.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Vectors & LLMs](<https://devfeed.tech/topics/vectors-llms.md>), [Python](<https://devfeed.tech/topics/python.md>), [spaCy](<https://devfeed.tech/topics/spacy.md>), [ann](<https://devfeed.tech/topics/ann.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [pip](<https://devfeed.tech/topics/pip.md>)

Tags: [1-5](<https://devfeed.tech/tags/1-5.md>), [2017](<https://devfeed.tech/tags/2017.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [algorithms](<https://devfeed.tech/tags/algorithms.md>), [animation](<https://devfeed.tech/tags/animation.md>), [api](<https://devfeed.tech/tags/api.md>), [application](<https://devfeed.tech/tags/application.md>), [applications](<https://devfeed.tech/tags/applications.md>), [approximate-nearest-neighbor-search](<https://devfeed.tech/tags/approximate-nearest-neighbor-search.md>), [context](<https://devfeed.tech/tags/context.md>), [databases](<https://devfeed.tech/tags/databases.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [filter](<https://devfeed.tech/tags/filter.md>), [llm](<https://devfeed.tech/tags/llm.md>), [python](<https://devfeed.tech/tags/python.md>), [query](<https://devfeed.tech/tags/query.md>), [store](<https://devfeed.tech/tags/store.md>), [text](<https://devfeed.tech/tags/text.md>), [vector](<https://devfeed.tech/tags/vector.md>), [vector-databases](<https://devfeed.tech/tags/vector-databases.md>)

### AI overview

ChromaDB stores text embeddings and retrieves semantically similar records, with metadata filters to narrow results. The tutorial explains vectors, cosine similarity, and embeddings, then demonstrates ChromaDB 1.5 with Python 3.12+: a persistent client stores data on disk, while in-memory and server or cloud clients are also available. A collection uses an embedding function such as SentenceTransformers' all-MiniLM-L6-v2 model and an HNSW index configured for cosine distance; queries return ranked matches, where cosine distance equals one minus cosine similarity. Examples cover adding documents with IDs and metadata, querying several prompts, filtering by metadata, and updating, upserting, or deleting records. The RAG example retrieves car reviews as context for an OpenAI LLM. The text cautions that semantic matches may be irrelevant when the collection or embedding model is a poor fit. Its supplied excerpt ends while describing batched ingestion of a large review dataset.

### Source excerpt

Learn what embeddings and vector databases are, then store, query, and filter text with ChromaDB in Python and use it to give an LLM context.

## OlmoEarth v1.3: State-of-the-art embeddings for the whole planet

DevFeed: [OlmoEarth v1.3: State-of-the-art embeddings for the whole planet](<https://devfeed.tech/articles/olmoearth-v1-3-state-of-the-art-embeddings-for-the-whole-planet-95941.md>)

Original publisher: [Read original article](<https://huggingface.co/blog/allenai/olmoearth-v1-3>)

Author: Kyle Wiggers

Published: 2026-10-22T16:11:31Z

Content type: release

Language: en

Sources: [Hugging Face](<https://devfeed.tech/sources/hugging-face-blog.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>)

Tags: [2017](<https://devfeed.tech/tags/2017.md>), [browser](<https://devfeed.tech/tags/browser.md>), [code](<https://devfeed.tech/tags/code.md>), [data](<https://devfeed.tech/tags/data.md>), [dimensions](<https://devfeed.tech/tags/dimensions.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embedding-model](<https://devfeed.tech/tags/embedding-model.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [explore](<https://devfeed.tech/tags/explore.md>), [explorer](<https://devfeed.tech/tags/explorer.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [maps](<https://devfeed.tech/tags/maps.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [olmoearth](<https://devfeed.tech/tags/olmoearth.md>), [report](<https://devfeed.tech/tags/report.md>), [satellite](<https://devfeed.tech/tags/satellite.md>), [story](<https://devfeed.tech/tags/story.md>), [tech](<https://devfeed.tech/tags/tech.md>), [tool](<https://devfeed.tech/tags/tool.md>)

### AI overview

OlmoEarth v1.3 turns annual Sentinel-1, Sentinel-2, and Landsat time series into 10-meter pixel embeddings in 128 dimensions, with a smaller 64-dimensional option. Precomputed embeddings are shared for reuse: train a lightweight classifier on them for mapping tasks, or compare yearly vectors to inspect change without training a model. Across 27 evaluations, the 128-dimensional version scored highest on 20 and ranked highest overall; the 64-dimensional version led on 18. The browser-based Embeddings Explorer lets users compare years, inspect Sentinel-2 imagery, find similar places, train a classifier from labeled points, and export GeoTIFFs without an account or GPU. Coverage is uneven: 2023-2025 are global, while 2017-2022 cover selected countries. Model weights, pretraining data, and code are open; Studio can generate embeddings for chosen dates, modalities, and resolution when the archive does not meet a project's needs.

### Source excerpt

🤗 Models | 📄 Tech Report | 📊 Data | 💻 Code | 🌎 Embeddings Explorer | 🗺 Interactive story Today, we're releasing OlmoEarth v1.3, a state-of-the-art satellite embedding model. Alongside OlmoEarth v1.3, we're sharing precomputed, annual embeddings for the whole planet (10m, 64- and 128-dimensions), plus a new browser-based tool to explore and customize maps based on those embeddings -- the OlmoEarth Embeddings Explorer.

## Ai2 releases OlmoEarth v1.3 state-of-the-art AI-ready satellite embeddings for the whole planet

DevFeed: [Ai2 releases OlmoEarth v1.3 state-of-the-art AI-ready satellite embeddings for the whole planet](<https://devfeed.tech/articles/ai2-releases-olmoearth-v1-3-state-of-the-art-ai-ready-satellite-embeddings-for-the-whole-planet-95741.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/10/22/ai2-releases-olmoearth-v1-3-state-of-the-art-ai-ready-satellite-embeddings-for-the-whole-planet/>)

Author: Kyt Dotson

Published: 2026-10-22T13:00:30Z

Content type: release

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI-Ready Data](<https://devfeed.tech/topics/ai-ready-data.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [1-3](<https://devfeed.tech/tags/1-3.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai2](<https://devfeed.tech/tags/ai2.md>), [allen-institute-for-ai](<https://devfeed.tech/tags/allen-institute-for-ai.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analyst](<https://devfeed.tech/tags/analyst.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [browser](<https://devfeed.tech/tags/browser.md>), [carbon](<https://devfeed.tech/tags/carbon.md>), [change](<https://devfeed.tech/tags/change.md>), [changes](<https://devfeed.tech/tags/changes.md>), [earth](<https://devfeed.tech/tags/earth.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embedding-model](<https://devfeed.tech/tags/embedding-model.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [fingerprint](<https://devfeed.tech/tags/fingerprint.md>), [model](<https://devfeed.tech/tags/model.md>), [news](<https://devfeed.tech/tags/news.md>), [olmoearth](<https://devfeed.tech/tags/olmoearth.md>), [radar-satellites](<https://devfeed.tech/tags/radar-satellites.md>), [releases](<https://devfeed.tech/tags/releases.md>), [resolutions](<https://devfeed.tech/tags/resolutions.md>), [satellite](<https://devfeed.tech/tags/satellite.md>), [satellite-ai-model](<https://devfeed.tech/tags/satellite-ai-model.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [top-story-6](<https://devfeed.tech/tags/top-story-6.md>)

### AI overview

Ai2's OlmoEarth v1.3 release pairs precomputed annual satellite embeddings for the planet at 10-, 64- and 128-meter resolutions with a browser-based explorer for making maps without signing in or using a GPU. Each embedding compresses a year of radar and visible/infrared observations into a numerical fingerprint that can help identify land-cover patterns and compare change over time. The excerpt contrasts this quick first-pass analysis with the prior workflow of fine-tuning a model for each task, which required substantial GPU resources and imagery storage; fine-tuning remains useful when deeper modeling is needed. Ai2 says v1.3 took 53% of v1.2's training compute time and 18% of v1's, while its carbon emissions were 50% and 22%, respectively. The article illustrates change analysis with reported mining-area comparisons in Peru, but provides no independent validation details.

### Source excerpt

Ai2, the Allen Institute for Artificial Intelligence, today released a satellite map embedding model called OlmoEarth v1.3 and precomputed annual embeddings for the whole planet at various resolutions with a browser-based map-making tool. The model compresses a year of satellite observations into a numerical fingerprint for each 10-by-10-meter patch of ground at the smallest resolution, [...] The post Ai2 releases OlmoEarth v1.3 state-of-the-art AI-ready satellite embeddings for the whole planet appeared first on SiliconANGLE.

## OlmoEarth v1.3: State-of-the-art embeddings for the whole planet

DevFeed: [OlmoEarth v1.3: State-of-the-art embeddings for the whole planet](<https://devfeed.tech/articles/olmoearth-v1-3-state-of-the-art-embeddings-for-the-whole-planet-95747.md>)

Original publisher: [Read original article](<https://allenai.org/blog/olmoearth-v1-3>)

Published: 2026-10-22T08:00:00Z

Content type: release

Language: en

Sources: [Ai2 Blog](<https://devfeed.tech/sources/ai2-blog.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [vector embeddings in AI](<https://devfeed.tech/topics/vector-embeddings-in-ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai2](<https://devfeed.tech/tags/ai2.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [analyze](<https://devfeed.tech/tags/analyze.md>), [archive](<https://devfeed.tech/tags/archive.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [browser](<https://devfeed.tech/tags/browser.md>), [carbon](<https://devfeed.tech/tags/carbon.md>), [change](<https://devfeed.tech/tags/change.md>), [code](<https://devfeed.tech/tags/code.md>), [compute](<https://devfeed.tech/tags/compute.md>), [computing](<https://devfeed.tech/tags/computing.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [crop](<https://devfeed.tech/tags/crop.md>), [data](<https://devfeed.tech/tags/data.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [explorer](<https://devfeed.tech/tags/explorer.md>), [global](<https://devfeed.tech/tags/global.md>), [olmoearth](<https://devfeed.tech/tags/olmoearth.md>), [satellite](<https://devfeed.tech/tags/satellite.md>)

### AI overview

OlmoEarth v1.3 provides reusable 10-meter satellite embeddings for mapping and change analysis, with precomputed coverage shared for community use. It combines annual Sentinel-1 radar, Sentinel-2 and Landsat time series into 128-dimensional pixel vectors; the first 64 dimensions offer a smaller option. Across 27 evaluations, the 128-dimensional embeddings led on 20 and ranked highest overall, while the 64-dimensional version led on 18. Users can compare years without training a model, or train a small classifier from labeled points in the browser-based Embeddings Explorer and export a GeoTIFF. The Explorer needs no account or GPU; OlmoEarth Studio remains available to run the model on chosen dates, modalities and resolution, or to fine-tune. Coverage is global for 2023-2025, while 2017-2022 covers selected countries. Provenance tables link tiles to the satellite scenes used.

### Source excerpt

OlmoEarth v1.3 brings state-of-the-art satellite embeddings, precomputed global data, and a browser-based Embeddings Explorer for mapping and analyzing the planet.

[Next page](<https://devfeed.tech/topics/vector-embeddings-in-ai.md?cursor=WyIyMDI2LTEwLTIyVDA4OjAwOjAwKzAwOjAwIiwgIjI4YzViYTA0LTE1ZmYtNDE4ZS04YmViLWE4YmI5ZjJhMGY5NyJd>)