REVIEW 49 cited by
jina-embeddings-v3: Multilingual Embeddings With Task LoRA
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
We introduce jina-embeddings-v3, a novel text embedding model with 570 million parameters, achieves state-of-the-art performance on multilingual data and long-context retrieval tasks, supporting context lengths of up to 8192 tokens. The model includes a set of task-specific Low-Rank Adaptation (LoRA) adapters to generate high-quality embeddings for query-document retrieval, clustering, classification, and text matching. Evaluation on the MTEB benchmark shows that jina-embeddings-v3 outperforms the latest proprietary embeddings from OpenAI and Cohere on English tasks, while achieving superior performance compared to multilingual-e5-large-instruct across all multilingual tasks. With a default output dimension of 1024, users can flexibly reduce the embedding dimensions to as low as 32 without compromising performance, enabled by Matryoshka Representation Learning.
Forward citations
Cited by 49 Pith papers
-
Leveraging Semantic Triples for Private Document Generation with Local Differential Privacy Guarantees
DP-ST uses semantic triples and neighborhood-aware local differential privacy to generate coherent private documents with better privacy-utility trade-offs than prior methods.
-
Large Concept Models: Language Modeling in a Sentence Representation Space
A sentence-level language model trained to autoregressively predict SONAR sentence embeddings can summarize, expand, and generate text in unseen languages.
-
DistilVDR: A Compact End-to-End Visual Document Retriever via Dual-Student Distillation
A 524M visual document retriever distilled from both sides of an 8B teacher reaches 86.9% of the teacher's NDCG@5 and beats every reproduced sub-1B baseline.
-
Robust Context-Aware Detection of Malicious Instructions in Text
A query- and context-aware sentence detector with adversarial training provides a tunable utility-versus-security defense against indirect prompt injection in LLM agents.
-
RoguePrompt: Dual-Layer Encoding for Self-Reconstruction to Circumvent LLM Moderation
RoguePrompt, a Vigenère+ROT13 self-reconstruction jailbreak, achieves 70.18% execution@3 and 93.93% bypass@3 across GPT-4o, Claude 3 Opus, and Gemini 1.5 Pro on 313 StrongREJECT prompts.
-
CuraWeb: Joint Optimization of Quality, Redundancy, and Diversity for Web-Scale Pretraining Data
CuraWeb jointly optimizes cleaning, deduplication, and sampling to build a 2T-token pretraining corpus that outperforms FineWeb-Edu, DCLM, Nemotron-CC, and Dolma3 by 1.82% average at 200B tokens.
-
KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking
KaLM-Reranker-V1 uses encoder–decoder FBNL with Matryoshka pooling to match Qwen3-class reranking quality at substantially lower online cost.
-
With Argus Eyes: Assessing Retrieval Gaps via Uncertainty Scoring to Detect and Remedy Retrieval Blind Spots
An entity-level Retrieval Probability Score exposes systematic blind spots in dense retrievers, predicts them from embedding geometry, and targeted Wikipedia-context augmentation of flagged entities raises nDCG@5/10 o...
-
NeoAMT: Neologism-Aware Agentic Machine Translation with Reinforcement Learning
NeoAMT trains an RL agent with a Wiktionary toolkit, novel reward, and adaptive rollouts to translate sentences containing neologisms in 16 languages and 75 directions.
-
DEBATE: A Large-Scale Benchmark for Evaluating Opinion Dynamics in Role-Playing LLM Agents
Using 2,792 humans' real debates as ground truth, role-playing LLM agents show excessive opinion convergence and public-stance drift compared with humans.
-
Negative Matters: Multi-Granularity Hard-Negative Synthesis and Anchor-Token-Aware Pooling for Enhanced Text Embeddings
A new MTEB state-of-the-art for text embeddings is reported by combining multi-granularity LLM-generated hard negatives with curriculum training and an anchor-token-aware pooling method.
-
LLM-as-a-Judge for Privacy Evaluation? Exploring the Alignment of Human and LLM Perceptions of Privacy in Textual Data
Across 10 text datasets, 13 LLMs, and 677 human raters, LLM privacy ratings align closely with the average human rating, while inter-human agreement is low.
-
SAGE-HLS: Syntax-Aware AST-Guided LLM for High-Level Synthesis Code Generation
A fine-tuned 7B LLM with AST-guided prompting generates HLS code that is near 100% synthesizable and 75% functionally correct, according to the abstract.
-
HyReC: Exploring Hybrid-based Retriever for Chinese
HyReC unifies dense, lexicon, and learned word-segment retrieval into one model and reports improved C-MTEB retrieval scores for Chinese.
-
jina-embeddings-v4: Universal Embeddings for Multimodal Multilingual Retrieval
jina-embeddings-v4 is a multimodal embedding model built on Qwen2.5-VL with LoRA adapters and dual single/multi-vector outputs, and it introduces the Jina-VDR benchmark for visually rich retrieval.
-
Decoupling Reasoning and Knowledge Injection for In-Context Knowledge Editing
DecKER decouples reasoning from knowledge editing by planning with masked placeholders before retrieving edited facts, improving multi-hop QA accuracy after knowledge edits.
-
Judging Quality Across Languages: A Multilingual Approach to Pretraining Data Filtering with Language Models
JQL trains small multilingual quality scorers from LLM judgments and human annotations, and filtering pretraining data with them improves downstream multilingual model performance over heuristic baselines.
-
A Text-Based Recommender System that Leverages Explicit Affective State Preferences
A proposed affective-cognitive recommender is evaluated with AC descriptions extracted from the gold book's own review, making the test a text-retrieval task rather than a test of preference-based recommendation.
-
Give LLMs a Security Course: Securing Retrieval-Augmented Code Generation via Knowledge Injection
Injecting retrieved vulnerability root-cause and fix knowledge into RACG prompts improves generated-code security rates by 20% on average and by 22% to 32% under poisoning, while preserving functional correctness.
-
Video-MMLU: A Massive Multi-Discipline Lecture Understanding Benchmark
Video-MMLU: a 1,065-video lecture benchmark where most AI video models score 10-50%, but text-only models answer 40% of quiz questions without video.
-
Fuzzy Linkography: Automatic Graphical Summarization of Creative Activity Traces
Fuzzy linkography automatically converts sequences of creative activity into weighted linkographs using embedding-model similarity, demonstrated on three diverse domains.
-
Exploring the Security Threats of Knowledge Base Poisoning in Retrieval-Augmented Code Generation
Poisoning the knowledge base of a retrieval-augmented code generator with vulnerable snippets raises the vulnerability rate of the model's output, with the size of the rise depending on the retriever, the model, and t...
-
LUSIFER: Language Universal Space Integration for Enhanced Multilingual Embeddings with Large Language Models
A two-stage, English-only training method aligns XLM-R representations with Mistral-7B to produce multilingual embeddings, improving low-resource language scores by up to 22 points while losing about 10 points on English.
-
AIR-Bench: Automated Heterogeneous Information Retrieval Benchmark
AIR-Bench uses LLMs to generate retrieval test data over 69 datasets, 9 domains and 13 languages, and reports a 0.82 rank correlation between model rankings on its generated MS MARCO set and human-labeled MS MARCO.
-
Latent Trajectory Discrimination for AI-Generated Text Detection
A sliding-window, trajectory-difference contrastive learner beats six AI-text detectors on RAID, NYT-AI, and OpenReview reviews.
-
Efficient Code Embeddings from Code Generation Models
jina-code-embeddings, 0.5B and 1.5B models built from Qwen2.5-Coder with last-token pooling, achieve strong code retrieval scores competitive with far larger embedding models.
-
Mind the Gap: Conformative Decoding to Improve Output Diversity of Instruction-Tuned Large Language Models
Instruction-tuning reduces LLM output diversity, DPO causes the biggest drop, and conformative decoding, a log-probability mixture of instruct and base models, partly restores diversity while keeping quality.
-
VulCoCo: A Simple Yet Effective Method for Detecting Vulnerable Code Clones
VulCoCo retrieves candidate code clones with embeddings and validates them with an LLM, outperforming prior vulnerable-clone detectors on a new synthetic benchmark and finding real-world clones that led to 15 CVEs.
-
SemCSE: Semantic Contrastive Sentence Embeddings Using LLM-Generated Summaries For Scientific Abstracts
Using multiple LLM-generated summaries of the same abstract as positive pairs trains scientific text embeddings that beat citation-trained baselines on retrieval and clustering, while the new benchmark shares its trai...
-
KinyaColBERT: A Lexically Grounded Retrieval Model for Low-Resource Retrieval-Augmented Generation
KinyaColBERT, a morphology-aware two-tier ColBERT retriever, reports large MRR gains over multilingual baselines and commercial APIs on a new Kinyarwanda agricultural retrieval benchmark.
-
Team RAS in 9th ABAW Competition: Multimodal Compound Expression Recognition Approach
A six-modality zero-shot pipeline with CLIP, Qwen-VL, WavLM, Mamba, and new fusion/aggregation modules reports F1 scores of 46.95 (AffWild2), 49.02 (AFEW), and 34.85 (C-EXPR-DB) without target-domain fine-tuning.
-
Chunk Twice, Embed Once: A Systematic Study of Segmentation and Representation Trade-offs in Chemistry-Aware Retrieval-Augmented Generation
A systematic evaluation shows that recursive 100-token non-overlapping chunks and retrieval-tuned embeddings outperform fixed-size chunks and domain-specific models like SciBERT for chemistry retrieval, and it introdu...
-
Enhancing Traffic Accident Classifications: Application of NLP Methods for City Safety
NLP models trained on German accident reports outperform tabular-only models for accident classification, and LLM analysis suggests many fallback 'other' labels are parking accidents.
-
PD$^3$: A Project Duplication Detection Framework via Adapted Multi-Agent Debate
A multi-agent debate system with a round-robin tournament design retrieves the most relevant prior projects and produces duplication scores and qualitative feedback for proposed power-sector research projects.
-
University of Indonesia at SemEval-2025 Task 11: Evaluating State-of-the-Art Encoders for Multi-Label Emotion Detection
Frozen BGE embeddings with per-emotion prompts and a CatBoost classifier outperform fully fine-tuned mBERT/XLM-R and surpass prior decoder baselines on 28-language emotion detection.
-
LongMagpie: A Self-synthesis Method for Generating Large-scale Long-context Instructions
By prompting an aligned LLM with a document and the special token that precedes a user query, LongMagpie synthesizes long-context instruction data that outperforms prior datasets when used to fine-tune Llama-3-8B.
-
ELITE: Embedding-Less retrieval with Iterative Text Exploration
An LLM-driven lexical search loop with an importance-based sufficiency check beats embedding-based RAG baselines on long-context QA benchmarks while cutting storage and preparation cost dramatically.
-
Reconstructing Context: Evaluating Advanced Chunking Strategies for Retrieval-Augmented Generation
A comparative evaluation on NFCorpus and MSMarco shows contextual retrieval beats late chunking on retrieval metrics at higher computational cost, and neither consistently beats standard early chunking.
-
jina-clip-v2: Multilingual Multimodal Embeddings for Text and Images
jina-clip-v2, an 865M-parameter multilingual dual-encoder, outperforms prior CLIP-style models on text-only and crossmodal retrieval, and on visually rich document retrieval, while supporting flexible embedding dimensions.
-
Bilingual BSARD: Extending Statutory Article Retrieval to Dutch
The authors extend the French BSARD legal retrieval dataset to Dutch (bBSARD) and benchmark retrieval models, showing small fine-tuned language-specific models can outperform zero-shot proprietary embeddings.
-
Arctic-Embed 2.0: Multilingual Retrieval Without Compromise
Arctic-Embed 2.0 delivers open multilingual embedding models with competitive MTEB-R and CLEF retrieval scores and strong 256-dimension MRL compression, plus new evidence on cross-lingual transfer.
-
Large Language Models Can Self-Improve in Long-context Reasoning
LLMs can improve their own long-context reasoning by fine-tuning on outputs selected by Minimum Bayes Risk consensus, without external annotations.
-
Adaptive Retrieval Strategies for Biomedical Question Answering
A BioASQ 13B system that adapts retrieval, reranking, and evidence clustering by question type achieved top Phase A retrieval scores, yet its own tables show non-adaptive baselines frequently performing better.
-
A Framework for Deductive Semantic Content Analysis at Scale in Science Education Using Text Embeddings
A few-shot text embedding classification framework achieves high agreement with human coders (Cohen's Kappa 0.74-0.83) on a simulated exhaustive coding task over 2,899 physics education survey responses.
-
When Noisy Labels Meet Class Imbalance on Graphs: A Graph Augmentation Method with LLM and Pseudo Label
GraphALP combines LLM-generated minority node oversampling with confidence-weighted pseudo-labeling to improve node classification on graphs that have both class imbalance and label noise.
-
Maintaining MTEB: Towards Long Term Usability and Reproducibility of Embedding Benchmarks
The MTEB maintainers document their infrastructure for versioning and validating benchmark components, plus a zero-shot score that flags models trained on benchmark tasks.
-
Hard Negative Mining for Domain-Specific Retrieval in Enterprise Systems
A reranker fine-tuned on hard negatives selected by two cosine-distance criteria outperforms older negative sampling methods on enterprise and domain-specific retrieval benchmarks.
-
BEIR-NL: Zero-shot Information Retrieval Benchmark for the Dutch Language
BEIR-NL is a Dutch-translated version of the BEIR benchmark with evaluations showing BM25 remains competitive against multilingual dense models.
-
LLMs are Also Effective Embedding Models: An In-depth Overview
A structured survey of using decoder-only LLMs as text embedding models, covering prompting, fine-tuning, data construction, benchmarks, and open problems.
Discussion (0). Continue with ORCID to comment.