Should you replace your text-embedding pipeline with a large language model? We answer this with a controlled, cost-aware comparison of ten LLMs across six families and 26 embedding models (118M to 14B parameters) on 37 tasks spanning classification, semantic textual similarity (STS), clustering, pair classification, and retrieval.
arXiv:2607. 23507v1 Announce Type: cross Abstract: Choosing the right text embedding model is one of the most consequential -- and most frequently under-examined -- decisions in building a retrieval or search system, yet the model that tops a leaderboard is rarely the best choice for a given deployment.
CORE improves compositional reasoning in multimodal language models by distilling a cross‑attentive reranker’s fine‑grained judgments into the embedding model. It generates candidate lists across five compositional matching levels and trains with a Rank‑KL objective to replicate the reranker’s ranking. Experiments on COLA, SUGARCREPE++, and NEGBENCH show CORE‑RERANKER‑8B outperforms Jina‑Reranker by 10.7 points, while CORE‑EMBED‑8B achieves the best overall average among evaluated embeddings, with gains also transferring to the MCMR benchmark without harming COCO or Flickr30K retrieval.
arXiv:2605. 11374v5 Announce Type: replace Abstract: Test-time compute is widely believed to benefit only large reasoning models, leaving small models with nothing to gain.
arXiv:2606.13061v3 Announce Type: replace Abstract: Reasoning-driven universal multimodal embedding has advanced rapidly by introducing Chain-of-Thought (CoT) reasoning into the embedding pipeline. D...
arXiv:2512. 10092v2 Announce Type: replace Abstract: Analyzing large-scale text corpora is a core challenge in machine learning, crucial for tasks like identifying undesirable model behaviors or biases in training data.
The paper investigates whether reasoning always benefits universal multimodal embeddings (UMEs). By comparing the discriminative and reasoning-driven branches of UME-R1, the authors find that while reasoning improves positive similarity in 56.6% of cases, it also creates 15.7% false-helpful instances where hard negatives are drawn closer. Diagnostic analyses reveal that reasoning often de‑condenses retrieved neighborhoods and that chain‑of‑thought tokens encode evidence common to both positives and hard negatives. Based on these insights, the authors introduce SURE, a utility router that boosts UME-R1‑7B by 1.5 points and consistently improves other embedding models on MMEB‑V2 without retraining or extra VLM passes.
The paper investigates whether large language models (LLMs) can leverage frozen relational‑transformer embeddings by injecting them as soft tokens. Using a learned MLP projection and LoRA adaptation, the authors fine‑tune Qwen3.5‑4B on chain‑of‑thought reasoning traces and group‑based reinforcement learning, then evaluate on ten binary classification tasks across six RelBench databases. The hybrid approach consistently underperforms the standalone relational transformer, showing sensitivity to serialization format, token budget, and RL stability, leading the authors to conclude that stronger alignment objectives and schema‑aware design are needed for reliable relational prediction.
MoEMB introduces a mixture‑of‑experts (MoE) approach to scale universal multimodal embeddings (UME) without increasing the size of the output vector or relying on autoregressive decoding. By expanding encoder capacity along the expert axis, MoEMB achieves state‑of‑the‑art performance on MMEB‑V2 and MRMR benchmarks with only 3 B active parameters, outperforming TTE‑based methods that use more than four times as many active parameters and require significantly more compute. The paper also presents the first comprehensive study of adaptive computation for MoE‑based embeddings, exploring training‑time and inference‑time strategies to further improve efficiency for large‑scale retrieval and recommendation systems.
The paper presents a two-level framework for scalable trade‑up recommendation. Level 1 distills large‑language‑model reasoning into a compact, non‑generative student that classifies product pairs using only precomputed embeddings, achieving high AUC on a benchmark. Level 2 applies product‑type test‑time training to fine‑tune lightweight adapters, further improving performance while keeping inference fast and inexpensive.
The paper introduces DEPT, a method that trains a single decoder-only large language model to both expand queries and encode documents for retrieval. By preserving document embeddings close to their initial cached values while allowing gradients to flow through the generator, DEPT stabilizes retrieval targets and enables efficient index reuse and online hard‑negative mining. Experiments on the BEIR benchmark with Qwen3‑4B‑Instruct‑2507 and LLaMA‑3.2‑3B‑Instruct show that DEPT outperforms training‑free, independently trained, and staged unified baselines, with ablations confirming the benefits of preservation, whitening, end‑to‑end expansion training, and online negatives.
arXiv:2606. 29532v1 Announce Type: cross Abstract: Integrating unstructured data into relational database systems is increasingly important as demand grows for natural language querying and analysis.
arXiv:2510. 20535v2 Announce Type: replace-cross Abstract: Recent techniques such as retrieval-augmented generation or chain-of-thought reasoning have led to longer contexts and increased inference costs.