ALEE generates AMR-based English minimal pairs with fine-grained semantic shifts, translates them, and evaluates embedding models on 275+ languages to expose cross-lingual gaps linked to training data and tokenization.
hub Mixed citations
Multilingual E5 Text Embeddings: A Technical Report
Mixed citation behavior. Most common role is method (43%).
82
Pith papers citing it
35
external citations · Pith
Method
43% of classified citations
abstract
This technical report presents the training methodology and evaluation results of the open-source multilingual E5 text embedding models, released in mid-2023. Three embedding models of different sizes (small / base / large) are provided, offering a balance between the inference efficiency and embedding quality. The training procedure adheres to the English E5 model recipe, involving contrastive pre-training on 1 billion multilingual text pairs, followed by fine-tuning on a combination of labeled datasets. Additionally, we introduce a new instruction-tuned embedding model, whose performance is on par with state-of-the-art, English-only models of similar sizes. Information regarding the model release can be found at https://github.com/microsoft/unilm/tree/master/e5 .
hub tools
citation-role summary
method 6
baseline 4
background 2
dataset 1
other 1
citation-polarity summary
representative citing papers
MMed-Bench-IR is a new heterogeneous benchmark spanning 6 languages and three non-overlapping tasks that exposes severe cross-lingual drops in biomedical retrieval performance.
HAKARI-Bench reconstructs 35 benchmarks into 551 tasks across 43 languages, reproducing full MTEB, MMTEB, and BEIR rankings with Spearman correlation above 0.97 while supporting efficiency variant comparisons.
Existing embedding models drop sharply from traditional code search to requirement-driven repository retrieval, while in-domain PR fine-tuning substantially improves Level-2/3 performance on CORE-Bench.
SEA-Embedding is a fully open text embedding pipeline for Southeast Asian languages that achieves state-of-the-art performance on the SEA-BED benchmark by analyzing data composition, training objectives, and base encoder choices.
HTEB introduces dynamic, multi-axis evaluation of text embedding robustness using LLM transformations, finding decoupled profiles across models and that scaling does not close all robustness gaps.
IdioLink introduces a benchmark dataset and evaluation showing that strong embedding models struggle to retrieve equivalent meanings across idiomatic and literal forms, relying on shallow cues instead.
Co-citation predictability for statute retrieval decays over 20 years in Ukrainian court data, dropping 33-47% in MRR with non-uniform patterns across legal domains.
DAPRO adaptively spends a fixed interaction budget across multi-turn LLM conversations and builds valid lower bounds on time-to-jailbreak, with coverage error scaling as the square root of the mean censoring weight.
EPIC trains LLMs to treat continuous embeddings as in-context prompts, yielding state-of-the-art text embedding performance on MTEB with or without prompts at inference and lower compute.
Defines ATIR task and benchmark for mixed audio-text queries; MLLM model with token compression shows substantial gains over strong baselines.
Code-switching creates a fundamental performance bottleneck for multilingual retrievers, causing drops of up to 27% on new benchmarks CSR-L and CS-MTEB, with embedding divergence as the key cause and vocabulary expansion insufficient to fix it.
Claim2Vec is a contrastively fine-tuned multilingual encoder that improves claim clustering performance and embedding space structure on multilingual fact-check datasets.
MultiSynt/MT supplies 4.8 trillion translated tokens in 36 languages from 100B English tokens, letting LLMs match native-data baselines with 72% fewer tokens and beat them by 15% at equal budget.
A Feedback Network model is developed showing online semantic exploration is more concentrated than physical mobility, with stable retail-business linkages and greater COVID disruption to spatial than cognitive routines, as a step toward hybrid digital twins of society.
BITEMBED trains 1.58-bit ternary-weight LLM embedders with contrastive pre-training, supervised distillation, and multi-precision output training, matching FP16 teachers within ~0.6 MMTEB points at ~2x CPU speed.
EvoEmbedding generates evolvable embeddings via a latent memory updated during sequential processing, outperforming larger models on long-context retrieval and generalizing to 10x longer contexts in downstream tasks.
Proposes a pretrained Universal Row Encoder using transformers and global statistics to generate table-width invariant row embeddings for modular relational graph models, claiming improved transfer, convergence, and memory on RelBench.
ARIADNE routes queries to the best adapter via embedding-space centroid proximity, recovering 97.44% of upper-bound performance on 23 NLP tasks and 89.7% selection accuracy on 44 tasks without training or internal access.
HyGRAG is a hierarchical graph RAG framework that constructs LLM summaries over hybrid chunk-entity graphs, retrieves via context and relation awareness across levels, and enables dynamic updates, reporting a 9.7% average accuracy gain on multi-hop reasoning tasks.
BBLP uses a multi-modal object encoder for label propagation in object detection and reaches 81.6% of fully-supervised mAP on D4LA with only 10% labelled data.
A multi-dimensional taxonomy filtering approach recovers high-performing data from deprioritized web corpora, with filtered low-tier subsets outperforming unfiltered top-tier data on reasoning and coding benchmarks.
FIGMA proposes a multi-view contrastive architecture plus the FGMCaps dataset to retrieve music from fine-grained textual descriptions of musical attributes, reporting up to 73.3% relative gains over CLAP baselines.
StoryVideoQA provides the largest auto-generated deep video understanding dataset to date with 363K QAs across TV and movies, paired with the PlotTree agent for hierarchical plot-based reasoning that existing VideoQA models struggle to match.
citing papers explorer
-
ALEE: Any-Language Evaluation of Embeddings via English-Centric Minimal Pairs
ALEE generates AMR-based English minimal pairs with fine-grained semantic shifts, translates them, and evaluates embedding models on 275+ languages to expose cross-lingual gaps linked to training data and tokenization.
-
MMed-Bench-IR: A Heterogeneous Benchmark for Multilingual Medical Information Retrieval
MMed-Bench-IR is a new heterogeneous benchmark spanning 6 languages and three non-overlapping tasks that exposes severe cross-lingual drops in biomedical retrieval performance.
-
HAKARI-Bench: A Lightweight Benchmark for Comparing Retrieval Architectures and Efficiency Settings under Unified Conditions
HAKARI-Bench reconstructs 35 benchmarks into 551 tasks across 43 languages, reproducing full MTEB, MMTEB, and BEIR rankings with Spearman correlation above 0.97 while supporting efficiency variant comparisons.
-
CORE-Bench: A Comprehensive Benchmark for Code Retrieval in the Era of Agentic Coding
Existing embedding models drop sharply from traditional code search to requirement-driven repository retrieval, while in-domain PR fine-tuning substantially improves Level-2/3 performance on CORE-Bench.
-
SEA-Embedding: Open and Reproducible Text Embeddings for Southeast Asia
SEA-Embedding is a fully open text embedding pipeline for Southeast Asian languages that achieves state-of-the-art performance on the SEA-BED benchmark by analyzing data composition, training objectives, and base encoder choices.
-
The Harder Text Embedding Benchmark (HTEB): Beyond One-dimensional Static Robustness
HTEB introduces dynamic, multi-axis evaluation of text embedding robustness using LLM transformations, finding decoupled profiles across models and that scaling does not close all robustness gaps.
-
IdioLink: Retrieving Meaning Beyond Words Across Idiomatic and Literal Expressions
IdioLink introduces a benchmark dataset and evaluation showing that strong embedding models struggle to retrieve equivalent meanings across idiomatic and literal forms, relying on shallow cues instead.
-
Temporal Decay of Co-Citation Predictability: A 20-Year Statute Retrieval Benchmark from 396M Ukrainian Court Citations
Co-citation predictability for statute retrieval decays over 20 years in Ukrainian court data, dropping 33-47% in MRR with non-uniform patterns across legal domains.
-
How Many Iterations to Jailbreak? Dynamic Budget Allocation for Multi-Turn LLM Evaluation
DAPRO adaptively spends a fixed interaction budget across multi-turn LLM conversations and builds valid lower bounds on time-to-jailbreak, with coverage error scaling as the square root of the mean censoring weight.
-
Embedding-based In-Context Prompt Training for Enhancing LLMs as Text Encoders
EPIC trains LLMs to treat continuous embeddings as in-context prompts, yielding state-of-the-art text embedding performance on MTEB with or without prompts at inference and lower compute.
-
ATIR: Towards Audio-Text Interleaved Contextual Retrieval
Defines ATIR task and benchmark for mixed audio-text queries; MLLM model with token compression shows substantial gains over strong baselines.
-
Code-Switching Information Retrieval: Benchmarks, Analysis, and the Limits of Current Retrievers
Code-switching creates a fundamental performance bottleneck for multilingual retrievers, causing drops of up to 27% on new benchmarks CSR-L and CS-MTEB, with embedding divergence as the key cause and vocabulary expansion insufficient to fix it.
-
Claim2Vec: Embedding Fact-Check Claims for Multilingual Similarity and Clustering
Claim2Vec is a contrastively fine-tuned multilingual encoder that improves claim clustering performance and embedding space structure on multilingual fact-check datasets.
-
MultiSynt/MT: Trillion-Token Multi-Parallel Pre-Training Data Translated Across 36 Languages
MultiSynt/MT supplies 4.8 trillion translated tokens in 36 languages from 100B English tokens, letting LLMs match native-data baselines with 72% fewer tokens and beat them by 15% at equal budget.
-
Toward a Hybrid Digital Twin of Society: Quantifying Cognitive-Spatial Linkages Through Online-Offline Feedback Networks
A Feedback Network model is developed showing online semantic exploration is more concentrated than physical mobility, with stable retail-business linkages and greater COVID disruption to spatial than cognitive routines, as a step toward hybrid digital twins of society.
-
BitNet Text Embeddings
BITEMBED trains 1.58-bit ternary-weight LLM embedders with contrastive pre-training, supervised distillation, and multi-precision output training, matching FP16 teachers within ~0.6 MMTEB points at ~2x CPU speed.
-
EvoEmbedding: Evolvable Representations for Long-Context Retrieval and Agentic Memory
EvoEmbedding generates evolvable embeddings via a latent memory updated during sequential processing, outperforming larger models on long-context retrieval and generalizing to 10x longer contexts in downstream tasks.
-
Universal Encoders for Modular Relational Deep Learning
Proposes a pretrained Universal Row Encoder using transformers and global statistics to generate table-width invariant row embeddings for modular relational graph models, claiming improved transfer, convergence, and memory on RelBench.
-
ARIADNE: Agnostic Routing for Inference-time Adapter DyNamic sElection
ARIADNE routes queries to the best adapter via embedding-space centroid proximity, recovering 97.44% of upper-bound performance on 23 NLP tasks and 89.7% selection accuracy on 44 tasks without training or internal access.
-
A Unified Framework for Context-Aware and Relation-Aware Graph Retrieval-Augmented Generation
HyGRAG is a hierarchical graph RAG framework that constructs LLM summaries over hybrid chunk-entity graphs, retrieves via context and relation awareness across levels, and enables dynamic updates, reporting a 9.7% average accuracy gain on multi-hop reasoning tasks.
-
Bounding Box Label Propagation for Re-Annotation of Document Layout Analysis Datasets
BBLP uses a multi-modal object encoder for label propagation in object detection and reaches 81.6% of fully-supervised mAP on D4LA with only 10% labelled data.
-
Unlocking Latent Value: Taxonomy-Guided Recovery of High-Performing Data from Low-Tier Web Corpora
A multi-dimensional taxonomy filtering approach recovers high-performing data from deprioritized web corpora, with filtered low-tier subsets outperforming unfiltered top-tier data on reasoning and coding benchmarks.
-
FIGMA: Towards FIne-Grained Music retrievAl
FIGMA proposes a multi-view contrastive architecture plus the FGMCaps dataset to retrieve music from fine-grained textual descriptions of musical attributes, reporting up to 73.3% relative gains over CLAP baselines.
-
StoryVideoQA: Scaling Deep Video Understanding with a Large-Scale, Multi-Genre and Auto-Generated Dataset
StoryVideoQA provides the largest auto-generated deep video understanding dataset to date with 363K QAs across TV and movies, paired with the PlotTree agent for hierarchical plot-based reasoning that existing VideoQA models struggle to match.
-
Boosting Self-Consistency with Ranking
RISC reformulates self-consistency answer selection as a ranking task solved by a lightweight LambdaRank model with five hand-designed features, yielding better accuracy-efficiency trade-offs than majority voting on QA benchmarks.
-
MIMO: Multilingual Information Retrieval via Monolingual Objectives
MIMO is a two-stage distillation-plus-contrastive framework that anchors multilingual embeddings to a monolingual English space and outperforms prior cross-lingual baselines on MLIR and multi-monolingual benchmarks.
-
On the Robustness of Multilingual Text Embedding Rankings Across Learning Tasks, Languages, and Benchmark Datasets
Meta-study of MTEB rankings introduces dataset-composition and ranking-scheme robustness indicators and finds only a small subset of models stay consistently strong across tasks, languages, and evaluation variations.
-
Structure Retention in Embedding Spaces as a Predictor of Benchmark Performance
Embedding model performance on MTEB tasks correlates strongly with nearest-neighbor overlap and ICA magnitude differences in their embedding spaces.
-
Can Large Audio Language Models Ignore Multilingual Distractors? An Evaluation of Their Selective Auditory Attention Capabilities
Introduces the MUSA benchmark and evaluates LALMs showing that strong single-speaker performance fails to ensure robust selective attention under multilingual interference, with errors from source confusion and unresolved attribution after separation.
-
An Annotation Scheme and Classifier for Personal Facts in Dialogue
An extended annotation scheme with new categories and attributes plus a Gemma-300M-based multi-head classifier achieves 81.6% macro F1 on personal fact classification, outperforming few-shot LLM baselines by nearly 9 points with lower compute.
-
jina-embeddings-v5-omni: Geometry-preserving Embeddings via Locked Aligned Towers
GELATO extends frozen Jina Embeddings v5 text models with locked non-text encoders, training only connectors to produce competitive multimodal embeddings while preserving exact text performance.
-
MLAIRE: Multilingual Language-Aware Information Retrieval Evaluation Protocal
MLAIRE is a protocol that evaluates multilingual retrievers on both semantic accuracy and query-language preference using parallel passages and new metrics like LPR and Lang-nDCG, showing that standard metrics hide distinct behavioral differences among retrievers.
-
Kernel Affine Hull Machines as Compute-Efficient Encoders for Frozen Semantic Spaces
KAHM yields a compute-efficient query encoder that outperforms matched learned adapters in reconstructing a frozen Mixedbread embedding space on an Austrian-law retrieval task while delivering an 8.53x CPU speedup.
-
Iterative Definition Refinement for Zero-Shot Classification via LLM-Based Semantic Prototype Optimization
Iterative LLM-based refinement of category definitions improves zero-shot classification performance across 13 embedding models on a new 10-category web URL benchmark.
-
RARE: Redundancy-Aware Retrieval Evaluation Framework for High-Similarity Corpora
RARE decomposes documents into atomic facts to build redundancy-aware retrieval benchmarks, revealing that strong retrievers drop from 66.4% to 5.0-27.9% PerfRecall@10 on high-overlap enterprise corpora versus standard benchmarks.
-
JFinTEB: Japanese Financial Text Embedding Benchmark
JFinTEB is the first benchmark for evaluating Japanese financial text embeddings across retrieval and classification tasks derived from realistic financial scenarios.
-
HIVE: Query, Hypothesize, Verify An LLM Framework for Multimodal Reasoning-Intensive Retrieval
HIVE raises multimodal retrieval nDCG@10 to 41.7 on the MM-BRIGHT benchmark by inserting LLM-driven hypothesis generation and verification between retrieval passes, delivering +9.5 over the best text-only baseline and +14.1 over the best multimodal baseline.
-
VERTIGO: Visual Preference Optimization for Cinematic Camera Trajectory Generation
VERTIGO post-trains camera trajectory generators with visual preference signals from Unity-rendered previews scored by a cinematically fine-tuned VLM, cutting character off-screen rates from 38% to near zero while improving framing and prompt adherence.
-
Learning to Retrieve from Agent Trajectories
Retrievers trained on agent trajectories via the LRAT framework improve evidence recall, task success, and efficiency in agentic search benchmarks.
-
LMEB: Long-horizon Memory Embedding Benchmark
LMEB is a new benchmark that evaluates embedding models on long-horizon memory retrieval and shows this skill is largely orthogonal to traditional passage-retrieval performance.
-
Adaptive Prompt Elicitation for Text-to-Image Generation
Adaptive Prompt Elicitation (APE) uses an information-theoretic framework to generate visual queries that elicit and compile user intent into better prompts for text-to-image models, showing improved alignment in benchmarks and a user study.
-
Reliable Evaluation Protocol for Low-Precision Retrieval
Proposes High-Precision Scoring (HPS) and Tie-aware Retrieval Metrics (TRM) to reduce tie-induced instability in low-precision retrieval evaluation.
-
Causal2Vec: Improving Decoder-only LLMs as Embedding Models through a Contextual Token
Causal2Vec prepends a BERT-generated contextual token to decoder-only LLMs and pools its hidden state with the EOS token to reach new SOTA on MTEB among public-data-trained embedding models.
-
Healthier LLMs: Retrieval-Augmented Generation for Public Health Question Answering
Hybrid RAG over UK public health guidance sharply raises MCQA accuracy and free-form faithfulness, letting smaller open models match larger closed models without retrieval.
-
Synergizing Zero-Shot Cross-Lingual Alzheimer Detection with Language-Invariant Multimodal Bi-Geometric Adversarial Learning
ORBIT framework fuses speech and text models with adversarial and bi-geometric learning to outperform baselines in zero-shot cross-lingual speech-based Alzheimer's detection.
-
Modeling semantic association in self-paced reading with language model embeddings
Sentence embeddings from language models capture semantic association effects on N400 and self-paced reading times beyond predictability, while word- and other context-based implementations do not.
-
DocRetriever: A Plug-and-Play Framework for Multimodal Document Retrieval with Comprehensive Benchmark
DocRetriever introduces a framework using layout-aware sparse embeddings for hybrid encoding without OCR and a generalizable reasoning-augmented reranker for few-shot settings, plus the MultiDocR benchmark for evaluation.
-
Adapting Multilingual Embedding Models to Turkish via Cross-Lingual Tokenizer Surgery and Offline Distillation
A 200M-parameter Turkish sentence embedding model is adapted from a multilingual teacher via tokenizer pruning, mean-composition initialization, and offline cosine distillation, achieving 77.55% Pearson correlation on STSbTR and 7th place on TR-MTEB.
-
Grounding Text Embeddings in Stakeholder Associations
The Stakeholder Grounding Exercise shows neural text embeddings are 19-26pp less reliable than human experts at capturing semantic distinctions, with misalignment strongly correlated to poorer clustering performance (ρ=0.9), replicated across Danish policy and US AI domains.
-
Lost in the Evidence? Reproducing Document Position and Context Size Effects in RAG
Reproducibility study shows position and context size effects in RAG depend on topic sampling and retrieval quality, proposes calibration for stable trends, and releases code after finding discrepancies with prior industry work.