Identifies the generative-discriminative gap in LLM hard negative synthesis for retrieval and proposes CausalNeg using CoT counterfactual perturbation plus query-view entropy maximization to generate more effective negatives.
hub
Gemini Embedding: Generalizable Embeddings from Gemini
32 Pith papers cite this work. Polarity classification is still indexing.
32
Pith papers citing it
abstract
In this report, we introduce Gemini Embedding, a state-of-the-art embedding model leveraging the power of Gemini, Google's most capable large language model. Capitalizing on Gemini's inherent multilingual and code understanding capabilities, Gemini Embedding produces highly generalizable embeddings for text spanning numerous languages and textual modalities. The representations generated by Gemini Embedding can be precomputed and applied to a variety of downstream tasks including classification, similarity, clustering, ranking, and retrieval. Evaluated on the Massive Multilingual Text Embedding Benchmark (MMTEB), which includes over one hundred tasks across 250+ languages, Gemini Embedding substantially outperforms prior state-of-the-art models, demonstrating considerable improvements in embedding quality. Achieving state-of-the-art performance across MMTEB's multilingual, English, and code benchmarks, our unified model demonstrates strong capabilities across a broad selection of tasks and surpasses specialized domain-specific models.
hub tools
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
SemaTune uses LLM guidance with semantic context to tune up to 41 Linux OS parameters, delivering 72.5% performance gains over defaults and 153.3% over non-LLM baselines on 13 workloads while avoiding degraded states.
TabEmbed is the first generalist embedding model for tabular data that unifies classification and retrieval in one space via contrastive learning and outperforms text embedding models on the new TabBench benchmark.
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.
Modern text encoders resist second-order collapse under mean pooling because token embeddings concentrate tightly within texts, and this resistance correlates with stronger downstream performance.
Semantic Recall is a new evaluation metric for approximate nearest neighbor search that focuses only on semantically relevant results, with Tolerant Recall as a proxy when relevance labels are unavailable.
Pico reduces LoRA merge interference by calibrating over-shared directions in the B matrix before merging, yielding 3.4-8.3 point accuracy gains and sometimes beating joint training.
LatentUMM proposes dual latent alignment at modality and capacity levels plus latent dynamics stabilization to reduce semantic drift and improve consistency in unified multimodal models.
Test-time LLM feedback refines query embeddings to deliver up to 25% relative gains on zero-shot literature search, intent detection, and related benchmarks.
Embeddings retrieve same-subfield papers at 45-52% but same-agenda papers at only 15-21%; citation rerank reaches 57-59% on agenda queries.
A survey that categorizes RIR benchmarks by domain and modality, proposes a taxonomy for integrating reasoning into retrieval pipelines, and outlines key challenges.
FLARE scores embedding models labellessly via normalized log-likelihood, achieving 0.90 Spearman correlation with supervised benchmarks and stable performance in dimensions over 3500 where prior methods collapse.
CLSGen is a dual-head LLM fine-tuning framework that enables joint probabilistic classification and verbalized explanation generation without catastrophic forgetting of generative capabilities.
A 300M-parameter open embedding model sets new SOTA on MTEB for its size class and matches models twice as large while staying effective when compressed.
ERA combines LLMs and tree search to produce expert-level empirical software that outperforms top human methods on single-cell analysis leaderboards and CDC COVID-19 forecasts.
Proposes High-Precision Scoring (HPS) and Tie-aware Retrieval Metrics (TRM) to reduce tie-induced instability in low-precision retrieval evaluation.
Controlled ablations of 38 models find MLM superior to CLM on representation benchmarks while CLM offers better data efficiency and stability; a biphasic CLM-then-MLM schedule is optimal under fixed compute and improves when initialized from pretrained CLM models.
LLMs exhibit 20-40% lower recall on ambiguous human names for PII detection, worsening under prompt injections, as shown via the new AmBench benchmark.
Combining dimensionality reduction and quantization compresses text embeddings to 0.1% size with minimal performance loss on MTEB tasks, outperforming either technique alone.
Three Metapath2Vec variants create ingredient embeddings by walking a co-occurrence graph from recipes, a typed chemical compound graph from FlavorDB, or a controlled blend of both.
LiSA improves AI guardrails lifelong by inducing conservative policies from sparse noisy failure reports via structured memory, conflict-aware rules, and posterior lower-bound gating.
EgoSelf uses graph-based memory of user interactions to derive personalized profiles and predict future behaviors for egocentric assistants.
Lack of exploration from conditioning on prior answers is the primary reason parallel sampling outperforms sequential sampling in large reasoning models.
Lightweight proxy models deliver over 100x cost and latency savings for semantic AI queries in databases with accuracy preserved or improved on benchmarks up to 10M rows.
citing papers explorer
-
When Hard Negatives Hurt: Bridging the Generative-Discriminative Gap in Hard Negative Synthesis for Retrieval
Identifies the generative-discriminative gap in LLM hard negative synthesis for retrieval and proposes CausalNeg using CoT counterfactual perturbation plus query-view entropy maximization to generate more effective negatives.
-
SemaTune: Semantic-Aware Online OS Tuning with Large Language Models
SemaTune uses LLM guidance with semantic context to tune up to 41 Linux OS parameters, delivering 72.5% performance gains over defaults and 153.3% over non-LLM baselines on 13 workloads while avoiding degraded states.
-
TabEmbed: Benchmarking and Learning Generalist Embeddings for Tabular Understanding
TabEmbed is the first generalist embedding model for tabular data that unifies classification and retrieval in one space via contrastive learning and outperforms text embedding models on the new TabBench benchmark.
-
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.
-
Why Mean Pooling Works: Quantifying Second-Order Collapse in Text Embeddings
Modern text encoders resist second-order collapse under mean pooling because token embeddings concentrate tightly within texts, and this resistance correlates with stronger downstream performance.
-
Semantic Recall for Vector Search
Semantic Recall is a new evaluation metric for approximate nearest neighbor search that focuses only on semantically relevant results, with Tolerant Recall as a proxy when relevance labels are unavailable.
-
Crowded in B-Space: Calibrating Shared Directions for LoRA Merging
Pico reduces LoRA merge interference by calibrating over-shared directions in the B matrix before merging, yielding 3.4-8.3 point accuracy gains and sometimes beating joint training.
-
LatentUMM: Dual Latent Alignment for Unified Multimodal Models
LatentUMM proposes dual latent alignment at modality and capacity levels plus latent dynamics stabilization to reduce semantic drift and improve consistency in unified multimodal models.
-
Task-Adaptive Embedding Refinement via Test-time LLM Guidance
Test-time LLM feedback refines query embeddings to deliver up to 25% relative gains on zero-shot literature search, intent detection, and related benchmarks.
-
Topic Is Not Agenda: A Citation-Community Audit of Text Embeddings
Embeddings retrieve same-subfield papers at 45-52% but same-agenda papers at only 15-21%; citation rerank reaches 57-59% on agenda queries.
-
A Survey of Reasoning-Intensive Retrieval: Progress and Challenges
A survey that categorizes RIR benchmarks by domain and modality, proposes a taxonomy for integrating reasoning into retrieval pipelines, and outlines key challenges.
-
FLARE: Task-agnostic embedding model evaluation through a normalization process
FLARE scores embedding models labellessly via normalized log-likelihood, achieving 0.90 Spearman correlation with supervised benchmarks and stable performance in dimensions over 3500 where prior methods collapse.
-
CLSGen: A Dual-Head Fine-Tuning Framework for Joint Probabilistic Classification and Verbalized Explanation
CLSGen is a dual-head LLM fine-tuning framework that enables joint probabilistic classification and verbalized explanation generation without catastrophic forgetting of generative capabilities.
-
EmbeddingGemma: Powerful and Lightweight Text Representations
A 300M-parameter open embedding model sets new SOTA on MTEB for its size class and matches models twice as large while staying effective when compressed.
-
An AI system to help scientists write expert-level empirical software
ERA combines LLMs and tree search to produce expert-level empirical software that outperforms top human methods on single-cell analysis leaderboards and CDC COVID-19 forecasts.
-
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.
-
Should We Still Pretrain Encoders with Masked Language Modeling?
Controlled ablations of 38 models find MLM superior to CLM on representation benchmarks while CLM offers better data efficiency and stability; a biphasic CLM-then-MLM schedule is optimal under fixed compute and improves when initialized from pretrained CLM models.
-
Can Large Language Models Really Recognize Your Name?
LLMs exhibit 20-40% lower recall on ambiguous human names for PII detection, worsening under prompt injections, as shown via the new AmBench benchmark.
-
When Is 0.1% Enough? Analyzing the Combined Effects of Dimensionality Reduction and Quantization on Text Embedding Compression
Combining dimensionality reduction and quantization compresses text embeddings to 0.1% size with minimal performance loss on MTEB tasks, outperforming either technique alone.
-
Epicure: Navigating the Emergent Geometry of Food Ingredient Embeddings
Three Metapath2Vec variants create ingredient embeddings by walking a co-occurrence graph from recipes, a typed chemical compound graph from FlavorDB, or a controlled blend of both.
-
LiSA: Lifelong Safety Adaptation via Conservative Policy Induction
LiSA improves AI guardrails lifelong by inducing conservative policies from sparse noisy failure reports via structured memory, conflict-aware rules, and posterior lower-bound gating.
-
EgoSelf: From Memory to Personalized Egocentric Assistant
EgoSelf uses graph-based memory of user interactions to derive personalized profiles and predict future behaviors for egocentric assistants.
-
Understanding Performance Gap Between Parallel and Sequential Sampling in Large Reasoning Models
Lack of exploration from conditioning on prior answers is the primary reason parallel sampling outperforms sequential sampling in large reasoning models.
-
100x Cost & Latency Reduction: Performance Analysis of AI Query Approximation using Lightweight Proxy Models
Lightweight proxy models deliver over 100x cost and latency savings for semantic AI queries in databases with accuracy preserved or improved on benchmarks up to 10M rows.
-
KScaNN: Scalable Approximate Nearest Neighbor Search on Kunpeng
KScaNN delivers up to 1.63x speedup on Kunpeng ARM over the best x86 ANNS solutions via hybrid intra-cluster search, improved PQ residuals, an ML adaptive module, and ARM-optimized SIMD kernels.
-
Retrofitting Small Multilingual Models for Retrieval: Matching 7B Performance with 300M Parameters
A 300M multilingual embedding model matches or exceeds 7B retrieval performance via optimized data scale, hard negatives, and task diversity over language diversity.
-
Improving Korean-English Cross-Lingual Retrieval: A Data-Centric Study of Language Composition and Model Merging
Language composition in training data creates opposing effects on CLIR and mono-IR performance for Korean-English retrieval, which model merging can partially resolve.
-
Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Ranking
Qwen3-VL-Embedding-8B achieves state-of-the-art performance with a 77.8 overall score on the MMEB-V2 multimodal embedding benchmark.
-
Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models
Qwen3 Embedding models in 0.6B-8B sizes achieve state-of-the-art results on MTEB and retrieval tasks including code, cross-lingual, and multilingual retrieval through unsupervised pre-training, supervised fine-tuning, and model merging on Qwen3 backbones.
-
Benchmarking LLMs on the Massive Sound Embedding Benchmark (MSEB)
LLMs exhibit a persistent modality gap versus specialized audio encoders on MSEB tasks, with no conclusive evidence favoring audio-native over cascaded architectures.
- FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings
- BLUEmed: Retrieval-Augmented Multi-Agent Debate for Clinical Error Detection