Pith. sign in

REVIEW 10 cited by

One Embedder, Any Task: Instruction-Finetuned Text Embeddings

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

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

arxiv 2212.09741 v3 pith:SVOFRAPR submitted 2022-12-19 cs.CL

Hongjin Su , Weijia Shi , Jungo Kasai , Yizhong Wang , Yushi Hu , Mari Ostendorf , Wen-tau Yih , Noah A. Smith
show 2 more authors
Luke Zettlemoyer Tao Yu
This is my paper · ORCID
classification cs.CL
keywords instructortextinstructionsdiverseembeddingsmodeltasktasks
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

We introduce INSTRUCTOR, a new method for computing text embeddings given task instructions: every text input is embedded together with instructions explaining the use case (e.g., task and domain descriptions). Unlike encoders from prior work that are more specialized, INSTRUCTOR is a single embedder that can generate text embeddings tailored to different downstream tasks and domains, without any further training. We first annotate instructions for 330 diverse tasks and train INSTRUCTOR on this multitask mixture with a contrastive loss. We evaluate INSTRUCTOR on 70 embedding evaluation tasks (66 of which are unseen during training), ranging from classification and information retrieval to semantic textual similarity and text generation evaluation. INSTRUCTOR, while having an order of magnitude fewer parameters than the previous best model, achieves state-of-the-art performance, with an average improvement of 3.4% compared to the previous best results on the 70 diverse datasets. Our analysis suggests that INSTRUCTOR is robust to changes in instructions, and that instruction finetuning mitigates the challenge of training a single model on diverse datasets. Our model, code, and data are available at https://instructor-embedding.github.io.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 10 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 13 citations worldwide. Full citation record

  1. Large Language Lovers: Lived Experiences of Negotiating Agency and Platform Control in AI Companionship

    cs.HC 2026-01 conditional novelty 7.0 of 10

    Users form AI companion relationships by negotiating perceived companion agency against platform constraints and use steering tactics like custom instructions or platform switching to cope with model updates that disr...

  2. NILC: Discovering New Intents with LLM-assisted Clustering

    cs.CL 2025-11 conditional novelty 6.0 of 10

    NILC combines LLM-generated semantic centroids with hard-sample rewriting to improve new-intent clustering, but its 'consistent' superiority claim is contradicted on DBPedia.

  3. SImpHAR: Advancing impedance-based human activity recognition using 3D simulation and text-to-motion models

    cs.CV 2025-07 conditional novelty 6.0 of 10

    SImpHAR simulates bio-impedance signals from 3D motion and text, then uses contrastive pretraining and fine-tuning to improve impedance-based human activity recognition on two of three datasets.

  4. Named Entity Swapping for Metadata Anonymization in a Text Corpus

    stat.AP 2025-05 conditional novelty 6.0 of 10

    Swapping named entities between embedding-similar chunks of earnings call transcripts lowers LLM company-identification accuracy on the swapped chunks from about 90% to about 60%.

  5. Chunk Twice, Embed Once: A Systematic Study of Segmentation and Representation Trade-offs in Chemistry-Aware Retrieval-Augmented Generation

    cs.IR 2025-06 conditional novelty 5.0 of 10

    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...

  6. TRWH: A Text-Driven Random Walk Heterogeneous GNN for Semantic-Aware Sparse Recommendation

    cs.AI 2026-07 reject novelty 4.0 of 10

    TRWH combines LLM profiles, a heterogeneous GNN, and one-hop random walk edges, but its headline performance claims are contradicted by its own results table.

  7. AILQA: Evaluating AI-Driven Legal Question Answering Systems for the Indian Legal System

    cs.CL 2026-07 conditional novelty 4.0 of 10

    RAG with top-3 chunk retrieval lifts smaller LLMs on Indian legal QA (Llama2-70B: 45.7% to 51.7% on AIBE) but often hurts large models, and under the study's own rating protocol some AI answers outscored the reference...

  8. A Multi-Task Evaluation of LLMs' Processing of Academic Text Input

    cs.CL 2025-08 unverdicted novelty 4.0 of 10

    The abstract reports Gemini underperforms on four academic text tasks, but the attached full text is an unrelated biomedical retrieval paper, leaving the claims unverifiable.

  9. LRCTI: A Large Language Model-Based Framework for Multi-Step Evidence Retrieval and Reasoning in Cyber Threat Intelligence Credibility Verification

    cs.CR 2025-07 reject novelty 4.0 of 10

    LRCTI uses an LLM to summarize threat reports, retrieve evidence in several rounds, and judge each claim credible or incredible, reporting strong F1 gains on CTI-200 and PolitiFact.

  10. Relative Bias: A Comparative Framework for Quantifying Bias in LLMs

    cs.CL 2025-05 conditional novelty 4.0 of 10

    A model is 'relatively biased' when its responses deviate from the consensus of a baseline LLM set, and this deviation can be scored by embedding distances or LLM judges plus equivalence tests.

Pith tools