AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Let LLMs Speak Embedding Languages: Generative Text Embeddings via Iterative Contrastive Refinement

Yu-Che Tsai, Kuan-Yu Chen, Yuan-Chi Li, Yuan-Hao Chen, Ching-Yu Tsai, Shou-De Lin

ICLR 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add soft-token generation (without stepwise loss or refinement) to Causal-EOS embedding (ablation)Mistral-7B, 50K samples · MTEB (English, v2)average score: 63.84 -> 65.21Improved
add stepwise contrastive loss to generative embedding (ablation)Mistral-7B, 50K samples · MTEB (English, v2)average score: 65.21 -> 65.69Improved
add iterative refinement regularization on top of stepwise loss (ablation)Mistral-7B, 50K samples · MTEB (English, v2)average score: 65.69 -> 66.27Improved
add soft-token generation to Causal-EOS embedding (ablation)Mistral-7B, 50K samples · Instruction following (IntentEmotion, NYT)average score: 47.05 -> 56.47Improved
add stepwise loss to generative embedding (ablation)Mistral-7B, 50K samples · Instruction followingaverage score: 56.47 -> 60.13Improved
add iterative refinement regularization (ablation)Mistral-7B, 50K samples · Instruction followingaverage score: 60.13 -> 62.97Improved
add iterative refinement regularization on top of stepwise loss (ablation)Mistral-7B, 50K samples · MTEB Summarizationscore: 35.28 -> 34.19Hurt
increase generated tokens at inference (K up to 20) though trained with K=5 (ablation)GIRCSE · MTEB STS, clustering, classificationrelative improvement over K=1Improved
generate more tokens at inference for non-generative Causal-EOS baseline (ablation)Causal-EOS · MTEB tasksscoreNo clear effect
use GIRCSE instead of Causal-EOS with 50K samples on Qwen2.5-0.5BQwen2.5-0.5B · MTEB averageaverage score: 55.5 -> 61.2Improved
use GIRCSE instead of Causal-EOS with 50K samples on Llama-3BLlama-3B · MTEB averageaverage score: 62.7 -> 65.5Improved
enable KV caching for GIRCSE inference (ablation)GIRCSE · FLOPs at sequence length 512, k=5FLOPs (T): 44.17 -> 7.40Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy