Instructions to use Wakeleyhuh/chowder-embed-specter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Wakeleyhuh/chowder-embed-specter with sentence-transformers:

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("Wakeleyhuh/chowder-embed-specter")

sentences = [
    "The hyperspace ω(f) when f is a transitive dendrite mapping",
    "Knee osteoarthritis (OA) is a major cause of chronic pain and disability in older adults, yet scalable home-based interventions remain limited, partly due to the lack of clinically effective digital solutions. This study is the first fully powered randomized, double-blind, sham-controlled trial to test a digitally delivered, home-based protocol combining transcranial direct current stimulation (tDCS) and mindfulness-based meditation (MBM) for knee OA pain. A total of 208 participants were randomized to active tDCS + MBM, active tDCS + sham MBM, sham tDCS + active MBM, or double sham; they completed ten 20-min sessions over 2 weeks. Knee pain intensity was measured at baseline, post treatment, and monthly for 3 months. Although both active tDCS groups improved, the difference between the combined tDCS + MBM intervention and tDCS alone was not statistically significant. Benefits were not sustained at 3 months. These findings offer the first evidence that a remotely supervised, digitally delivered tDCS + MBM intervention can rapidly reduce knee OA pain, supporting future personalized and extended treatment studies. Clinical trial registration: ClinicalTrials.gov NCT04375072 (registered May 7, 2020).",
    "  Let $X$ be a compact metric space. By $2^X$ we denote the hyperspace of all\nclosed and non-empty subsets of $X$ endowed with the Hausdorff metric. Let\n$f:X\\to X$ be a continuous function. In this paper we study some topological\nproperties of the hyperspace $\\omega(f)$, the collection of all omega limits\nsets $\\omega(x,f)$ with $x\\in X$. We prove the following: $i)$ If $X$ has no\nisolated points, then, for every continuous function $f:X\\to X$,\n$int_{2^X}(\\omega(f))=\\emptyset$. $ii)$ If $X$ is a dendrite for which every\narc contains a free arc and $f:X\\to X$ is transitive, then the hyperspace\n$\\omega(f)$ is totally disconnected. $iii)$ Let $D_\\infty$ be the Wazewski's\nuniversal dendrite. Then there exists a transitive continuous function\n$f:D_\\infty\\to D_\\infty$ for which the hyperspace $\\omega(f)$ contains an arc;\nhence, $\\omega(f)$ is not totally disconnected.\n",
    "Let $X_1,\\,X_2,\\,\\ldots,\\,X_N$, $N\\in\\mathbb N$ be independent, discrete, integer-valued random variables. Assume that $X_j\\geqslant m_j$ almost surely for each $j=1,\\,2,\\,\\ldots,\\,N$, where $m_1,\\,m_2,\\,\\ldots,\\,m_N\\in\\mathbb{Z}$ satisfy $m_1+\\cdots+m_N<0$. Furthermore, suppose that the sequence $X_1,\\,X_2,\\,\\ldots$ is periodic in distribution, i.e. $X_k{\\buildrel d \\over =} X_{k+N}$ for all $k\\in\\mathbb N$. We derive computable representations for the distribution functions of $\\max\\{X_1,\\,X_1+X_2,\\,\\ldots\\}$, $\\max\\{X_2,\\,X_2+X_3,\\,\\ldots\\}$, $\\ldots$, $\\max\\{X_N,\\,X_N+X_{N+1},\\,\\ldots\\}$. The obtained formulas are based on a linear recurrence whose initial values are determined from a linear system that involves the roots of an associated characteristic equation and the distributions of $X_1,\\,X_2,\\,\\ldots,\\,X_N$. Several examples are presented, including a biseasonal-biased Rademacher random walk for which the distribution, generating functions, and all moments admit explicit closed-form expressions. In addition, we identify and correct several inaccuracies in the results reported in \\cite{Grigutis2024}."
]
embeddings = model.encode(sentences)

similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [4, 4]

Notebooks
Google Colab
Kaggle

chowder-embed-specter / README.md

Wakeleyhuh

Add new SentenceTransformer model

8c06721 verified 1 day ago

preview code

Raw

History Blame Contribute Delete

73.8 kB

metadata

language:
  - en
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:70764
  - loss:MultipleNegativesRankingLoss
base_model: sentence-transformers/allenai-specter
widget:
  - source_sentence: The hyperspace ω(f) when f is a transitive dendrite mapping
    sentences:
      - >-
        Knee osteoarthritis (OA) is a major cause of chronic pain and disability
        in older adults, yet scalable home-based interventions remain limited,
        partly due to the lack of clinically effective digital solutions. This
        study is the first fully powered randomized, double-blind,
        sham-controlled trial to test a digitally delivered, home-based protocol
        combining transcranial direct current stimulation (tDCS) and
        mindfulness-based meditation (MBM) for knee OA pain. A total of 208
        participants were randomized to active tDCS + MBM, active tDCS + sham
        MBM, sham tDCS + active MBM, or double sham; they completed ten 20-min
        sessions over 2 weeks. Knee pain intensity was measured at baseline,
        post treatment, and monthly for 3 months. Although both active tDCS
        groups improved, the difference between the combined tDCS + MBM
        intervention and tDCS alone was not statistically significant. Benefits
        were not sustained at 3 months. These findings offer the first evidence
        that a remotely supervised, digitally delivered tDCS + MBM intervention
        can rapidly reduce knee OA pain, supporting future personalized and
        extended treatment studies. Clinical trial registration:
        ClinicalTrials.gov NCT04375072 (registered May 7, 2020).
      - >2
          Let $X$ be a compact metric space. By $2^X$ we denote the hyperspace of all
        closed and non-empty subsets of $X$ endowed with the Hausdorff metric.
        Let

        $f:X\to X$ be a continuous function. In this paper we study some
        topological

        properties of the hyperspace $\omega(f)$, the collection of all omega
        limits

        sets $\omega(x,f)$ with $x\in X$. We prove the following: $i)$ If $X$
        has no

        isolated points, then, for every continuous function $f:X\to X$,

        $int_{2^X}(\omega(f))=\emptyset$. $ii)$ If $X$ is a dendrite for which
        every

        arc contains a free arc and $f:X\to X$ is transitive, then the
        hyperspace

        $\omega(f)$ is totally disconnected. $iii)$ Let $D_\infty$ be the
        Wazewski's

        universal dendrite. Then there exists a transitive continuous function

        $f:D_\infty\to D_\infty$ for which the hyperspace $\omega(f)$ contains
        an arc;

        hence, $\omega(f)$ is not totally disconnected.
      - >-
        Let $X_1,\,X_2,\,\ldots,\,X_N$, $N\in\mathbb N$ be independent,
        discrete, integer-valued random variables. Assume that $X_j\geqslant
        m_j$ almost surely for each $j=1,\,2,\,\ldots,\,N$, where
        $m_1,\,m_2,\,\ldots,\,m_N\in\mathbb{Z}$ satisfy $m_1+\cdots+m_N<0$.
        Furthermore, suppose that the sequence $X_1,\,X_2,\,\ldots$ is periodic
        in distribution, i.e. $X_k{\buildrel d \over =} X_{k+N}$ for all
        $k\in\mathbb N$. We derive computable representations for the
        distribution functions of $\max\{X_1,\,X_1+X_2,\,\ldots\}$,
        $\max\{X_2,\,X_2+X_3,\,\ldots\}$, $\ldots$,
        $\max\{X_N,\,X_N+X_{N+1},\,\ldots\}$. The obtained formulas are based on
        a linear recurrence whose initial values are determined from a linear
        system that involves the roots of an associated characteristic equation
        and the distributions of $X_1,\,X_2,\,\ldots,\,X_N$. Several examples
        are presented, including a biseasonal-biased Rademacher random walk for
        which the distribution, generating functions, and all moments admit
        explicit closed-form expressions. In addition, we identify and correct
        several inaccuracies in the results reported in \cite{Grigutis2024}.
  - source_sentence: New mechanism for delayed ENSO impact on the East Asian Summer Monsoon
    sentences:
      - >-
        Abstract Due to its potentially life—threatening and devastating
        economic impacts, variability in the East Asian Summer Monsoon needs
        better understanding, more accurate simulation and improved prediction.
        Much of the current ability of long-range summer monsoon forecasts stems
        from the link to the preceding winter El Niño Southern Oscillation, but
        the mechanisms behind this lagged impact are not fully understood. In
        this study, a simple new mechanism is proposed, based on migrating
        Atmospheric Angular Momentum anomalies. The polewards migration of
        atmospheric angular momentum associated with winter El Niño is shown to
        impact the northwest Pacific anticyclone responsible for much of the
        year-to-year variability in the monsoon. Model forecasts of the summer
        monsoon are shown to have increased accuracy when this new mechanism is
        active, with the increase in the success rate of forecasts due to
        migrating angular momentum anomalies shown to be as large as the effect
        of El Niño itself.
      - >-
        Automated algorithm selection in black-box optimization typically relies
        on supervised models that map landscape features to algorithm
        performance labels. Such models are costly to train,
        benchmark-dependent, and often fail to generalize to unseen problem
        classes. We study an unsupervised alternative: multi-kernel clustering
        over heterogeneous landscape representations, in which problem instances
        are grouped without using performance labels in the clustering stage,
        and the resulting clusters are mapped post hoc to solver recommendations
        through a strictly separated three-stage evaluation protocol. Drawing on
        two decades of advances in multiple kernel learning, we adopt a
        multi-kernel k-means formulation that jointly learns cluster assignments
        and kernel weights over four heterogeneous landscape views: ELA,
        DeepELA, DoE2Vec, and TransOptAS. On affine BBOB-derived selector tasks
        for Differential Evolution (DE) and Particle Swarm Optimization (PSO) at
        a fixed evaluation budget, we report mean plus or minus standard
        deviation selector profiles over 50 independent random seeds for
        stochastic configurations. Multi-kernel clustering obtains the strongest
        mean profile on the DE portfolio and remains competitive with, and
        nominally ahead of, the leading baselines on the more compressed PSO
        portfolio, where differences among the best methods are small relative
        to stochastic variation. In representative median-seed runs used for
        visualization, the learned kernel weights retain ELA and TransOptAS
        while assigning zero weight to DeepELA and DoE2Vec, providing a
        task-specific interpretation of which representations are retained by
        the multi-kernel model for selector-oriented grouping.
      - >-
        Prolonged social isolation (SI) and negative emotion are associated with
        an increased risk of cardiovascular diseases (CVDs). However, it remains
        elusive whether SI and emotional states affect the pathological process
        of myocardial infarction (MI). In this study, SI models with different
        duration and MI model were co-established in mice. Anxiety and
        depression were assessed by a series of behavioral tests including open
        field test, elevated plus maze test, novelty-suppressed feeding test,
        tail suspension test and forced swim test. Cardiac function, heart
        infarct size and fibrosis were assessed by echocardiography, TTC
        staining, and Masson staining. The activity of neurons across the whole
        brain, as well as the hypothalamic-pituitary-adrenal (HPA) axis, was
        also investigated. Long-term SI induced anxiety and depression-like
        behaviors and aggravated cardiac injury and inflammatory response in MI
        mice, while short-term SI induced anxiety-like behavior but not
        depression-like behavior and had no significant effects on cardiac
        injury. Long-term SI altered the activity of several brain regions
        related to emotional, reward, autonomic and neuroendocrine regulation in
        MI mice. Furthermore, serum corticosterone levels were altered,
        indicating the potential involvement of HPA axis. These findings reveal
        that long-term SI exacerbated cardiac dysfunction and cardiac injury
        after MI. The underlying mechanisms involved in this process may include
        negative mood, dysregulation of the reward system, autonomic nervous
        system modulation, and HPA axis activation. These findings will
        contribute to our better understanding of heart-brain connections.
  - source_sentence: Towards a Theory of Dobrakov-Sobolev Spaces
    sentences:
      - >-
        The aim of this paper is to introduce a generalization of Sobolev spaces
        based on the Dobrakov integral. More precisely, we consider the setting
        of Banach-valued functions and Fomin differentiable Borel
        operator-valued measures on a finite-dimensional space. To build the
        necessary rigorous foundation, we establish analogs of several key
        results from the theory of differentiable real-valued measures,
        including the Leibniz rule and the integration by parts formula, all
        within the context of Dobrakov integration. These results are then
        embedded into the general scheme of vector-valued distribution theory.
        In particular, we describe the configuration of test spaces that yields
        an appropriate definition of a generalized derivative with respect to a
        differentiable operator-valued measure.
      - >-
        AI agents are increasingly used to automate research and development
        tasks, yet existing benchmarks typically evaluate them on prescribed
        workflows or narrow action spaces. Even nominally open-ended tasks can
        often be solved by retrieving a well-known recipe and tuning a few
        hyperparameters, making it unclear whether strong results reflect
        genuine optimization or memorized solutions. We introduce
        InferenceBench, where an agent must deploy an OpenAI-compatible
        inference server and optimize the speed of LLM inference. Each agent
        receives a target LLM, one H100 GPU, an optimization scenario, and a
        wall-clock time budget of two hours. Three optimization scenarios
        isolate distinct bottlenecks of inference (prefill latency, decode
        latency, and concurrent request throughput) and a fourth balances all
        three at the same time. Across 15 frontier agent configurations, agents
        reliably improve over a naive PyTorch baseline (up to $8.08\times$) and
        often match or exceed serving engines with default settings
        ($4.05\times$ for vLLM), but still fall below a simple hyperparameter
        search under the same time budget (up to $11.53\times$). Qualitative
        analysis of agent trajectories shows that although agents enumerate many
        relevant optimization techniques, they overwhelmingly converge on a
        single inference framework. They test only a few distinct configurations
        and spend the remaining budget re-measuring, repairing, or optimizing
        hyperparameters rather than exploring substantially different
        strategies. This suggests the bottleneck is not domain knowledge, but
        the ability to propose diverse configurations, evaluate them
        systematically, and submit the best identified solution. Overall,
        InferenceBench reflects the ability of agents to operate in an
        open-ended AI engineering setting, where memorized solutions lead to
        limited improvements.
      - >-
        Abstract Patients with multiple myeloma bearing a deletion of chromosome
        17p (del(17p)), mutation of TP53 , or both have poorer prognosis
        compared to patients without these aberrations. We investigated the
        activity and mechanism of melflufen (melphalan flufenamide) in myeloma
        models with wild type TP53 ( TP53 wt) and complete TP53 deletion ( TP53
        −/− ) and assessed the efficacy of melflufen in patients with del(17p)
        and/or TP53 mutation. Ex vivo data from myeloma plasma cells (PC) showed
        comparable activity of melflufen in del(17p), TP53 −/− , and TP53 wt
        samples. scRNAseq data demonstrated that melflufen sensitive PCs had
        lower expression of p53 target genes and higher expression of genes
        associated with DNA damage repair and cell cycle checkpoints.
        Irrespective of TP53 status, melflufen induced apoptosis, DNA damage,
        and mitochondrial dysfunction, while only in TP53 −/− cells, it led to
        changes in expression of cell cycle checkpoint and apoptosis genes.
        Post-hoc analysis of the OCEAN trial melflufen-treated del(17p) patient
        population also demonstrated favorable progression free survival
        compared to pomalidomide-treated cohort. Our insights into the molecular
        mechanisms of melflufen activity in TP53 −/− myeloma support its
        clinical efficacy and application in the del(17p) and TP53 −/− patient
        population. Trial registration NCT03151811, registration 2017-05-09.
  - source_sentence: >-
      Strengthening complexity in EU: the role of educational policy and R&amp;D
      spending
    sentences:
      - >-
        The drivers of economic complexity in the countries of the European
        Union (EU) are a question of significant importance, especially in the
        context of the emergence of a trade war. Studying the relationship
        between economic complexity and macroeconomic, educational and
        technological variables, we offer valuable conclusions for the design
        and formulation of new policies. The Economic Complexity Index (ECI) is
        analysed in the context of education expenditure, gross fixed capital
        formation, fiscal balance, R&D expenditure, and patent applications,
        using relevant econometric techniques such as the 3SLS and the SURE
        analysis. Education expenditures are found to affect significantly the
        Economic Complexity of a country. Moreover, after the outburst of the
        sovereign-debt crisis the fiscal balance has become a prerequisite for
        all European countries, showing a positive relationship with economic
        complexity. This result poses the question of how weaker economies can
        finance the necessary policies to increase their economic complexity,
        such as raising spending on education, R&D, and fixed capital
        investment, while maintaining fiscal stability.
      - >-
        Proxy metrics are widely used to improve the precision and velocity of
        online experimentation (aka A/B testing). Although proxies are often
        motivated by long-term outcomes that the experimenter does not observe,
        in many settings they are used alongside a contemporaneous but
        statistically insensitive north star. This can lead to a practical
        dilemma: when should experimenters trust the proxy metric, and when
        should they trust the north star? In this paper, I propose an optimal
        blending approach that smoothly guides decision-making towards the north
        star as the power of the experiment increases and away from the north
        star as the quality of the proxy metric improves. I study the
        implications of this decision-making framework for the design of
        experiments and of experimentation programs. Equipped with better
        (worse) proxy metrics, experimenters should run smaller and more (larger
        and fewer) experiments. I show how to leverage past experiments to
        estimate optimal blending weights and experiment sizes. Lastly, I
        describe the real-world application of the methodology to an
        experimentation program at Netflix.
      - >-
        We consider a quenched SU(2)$\times$U(1) gauge Higgs theory on the
        lattice, coupled to a static vector-like fermion which, in this case, is
        in the same gauge group representation as the Higgs field. Physical
        (i.e. locally gauge invariant) electrically charged and electrically
        neutral states of matter particles in the electroweak theory were
        described decades ago, but those constructions do not exhaust all the
        possibilities, and new types of electrically charged/neutral states,
        orthogonal to former constructions, are described here. The difference
        has to do with how the static source, which by itself does not create a
        physical state, is dressed by dynamical fields. We find that,
        unsurprisingly, the neutral static fermion is much lighter than any of
        the charged fermion states. But a lattice study of the propagation of
        the charged fermion states indicates the existence of (at least) two
        particle states with different masses in charged particle spectrum.
  - source_sentence: >-
      ERBB3 blockade sensitizes hepatocellular carcinoma to regorafenib after
      first-line tyrosine kinase inhibitor resistance by inhibiting HIF1A-ABCB1
      signaling
    sentences:
      - >-
        We improve the theoretical predictions for rare $Z$-boson decays, $Z\to
        V\ell^+\ell^-$ ($\ell=e$ or $\mu$), where $V$ denotes a heavy vector
        quarkonium including $J/\Psi$, $\Psi (2S)$, and $\Upsilon (nS)$ with
        $n=1,2,3$. These processes are thought to be dominated by the
        electromagnetic fragmentation transition, i.e., $Z\to
        \gamma^*\ell^+\ell^-$ followed by $\gamma^*\rightarrow V$. The present
        study includes all of the relevant tree-level Feynman diagrams, which
        contribute to these decays in the standard model. Our analysis shows
        that, for the charmonium final states, the fragmentation transition
        almost saturates the whole contribution and the other diagrams can be
        neglected; while for the bottomonium final states, the inclusion of
        other diagrams can increase their branching fractions by $4\%\sim 9\%$.
        Further investigation of the differential distributions, especially the
        angular distributions, indicates that forward-backward asymmetries for
        final leptons in these processes would be zero in the standard model.
        Therefore, in future experimental facilities with large number of
        $Z$-boson events accumulated, studies of these rare $Z$ decays may help
        both to test the standard model and to probe its interesting extensions.
      - >-
        Historically, veterinary studies screening for breed, age and sex
        predisposition to disease have relied on collating small-scale studies
        of clinical datasets. The availability of larger datasets through groups
        such as the Small Animal Veterinary Surveillance Network (SAVSNET)
        promise access to information regarding a wide range of clinical
        presentations at scale, however, methodological limitations surrounding
        the extraction of specific disease information or screening for disease
        predispositions result in a substantial reduction in the number of
        animals studied. These studies often address very focused hypotheses -
        only leveraging a small fraction of the intrinsic value of the data at
        any one time. Here, we implemented an unsupervised machine learning
        methodology, creating a representation of a large volume of clinical
        notes collected by SAVSNET from veterinary practices across the UK. We
        utilise BERTopic, a topic-modelling tool based on Bidirectional Encoder
        Representations using Transformers (BERT) architecture, and show it is
        able to surface known phenotypes, such as breed predispositions to
        hypoadrenocorticism, diabetes mellitus and mitral valve disease, as well
        as potential novel patterns of disease phenotypes. This scalable and
        granular modelling technique facilitates the rapid interrogation of
        large clinical datasets, enabling the identification of a broad range of
        phenotypes within the population and the early detection of temporal
        changes indicative of emerging infectious or environmental diseases.
        Supplementary Information: The online version contains supplementary
        material available at 10.1186/s40537-026-01365-0.
      - >-
        BACKGROUND/AIMS: Regorafenib is recommended by guidelines and trials as
        a sequential second-line therapy following progression on first-line
        sorafenib or lenvatinib in hepatocellular carcinoma (HCC). However,
        efficacy is limited, highlighting the urgent need to screen suitable
        patients and develop sensitization strategies. METHODS: Acquired
        sorafenib- or lenvatinib-resistant (SR or LR) HCC cell lines and
        organoids were established. Genome-wide CRISPR library screen was
        performed in SR or LR cell strains to identify synthetic lethal targets
        of regorafenib. RNA-seq and FITC-regorafenib efflux assay were used to
        elucidate ERBB3-driven downstream signaling. Preclinical mouse models of
        cell line- and patient-derived xenografts and clinical cohorts of HCC
        patients were employed to validate the efficacy of ERBB3-guided patient
        stratification. RESULTS: Screening with CRISPR library, we showed that
        inhibition of ERBB3 was synthetic lethal with regorafenib in SR or LR
        cell strains and organoids. Mechanistically, SR or LR triggered feedback
        activation of ERBB3 signaling and mediated regorafenib efflux via
        ERBB3-HIF1A-ABCB1 cascade pathway, limiting sensitivity to regorafenib.
        Moreover, ERBB3-low tumors following SR or LR exhibited significant
        sensitivity to regorafenib, suggesting its potential as a predictive
        biomarker to screen optimal candidates for sequential therapy.
        Seribantumab, an ERBB3-targeting monoclonal antibody, inhibited
        ERBB3-HIF1A-ABCB1 cascade, and its combination with regorafenib exerted
        marked synergistic anti-tumor effects on ERBB3-high tumors resistant to
        sorafenib or lenvatinib both in vitro and in vivo. CONCLUSIONS: This
        study revealed that ERBB3 was a key resistance factor driving limited
        efficacy to sequential regorafenib, but also an effective therapeutic
        target whose inhibition enhanced regorafenib sensitivity after SR or LR.
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
model-index:
  - name: chowder-embed (SPECTER fine-tuned on the chowder corpus)
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: NanoSCIDOCS
          type: NanoSCIDOCS
        metrics:
          - type: cosine_accuracy@1
            value: 0.38
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.62
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.72
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.86
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.38
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.31333333333333335
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.256
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.176
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.08066666666666666
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.19466666666666665
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.26266666666666666
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.3616666666666666
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.3456136739010105
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.534436507936508
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.26477117793177524
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: NanoSciFact
          type: NanoSciFact
        metrics:
          - type: cosine_accuracy@1
            value: 0.6
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.68
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.72
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.8
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.6
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.24666666666666665
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.16
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.565
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.67
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.715
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.79
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.6831330092247734
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.6557142857142857
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.6496795415115648
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: NanoNFCorpus
          type: NanoNFCorpus
        metrics:
          - type: cosine_accuracy@1
            value: 0.3
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.36
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.46
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.54
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.3
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.2533333333333333
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.248
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.222
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.012486154892961041
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.0309389720826588
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.047390511275329145
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.10215743728341801
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.24495153881224493
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.3605714285714285
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.09185350554401533
            name: Cosine Map@100
      - task:
          type: nano-beir
          name: Nano BEIR
        dataset:
          name: NanoBEIR mean
          type: NanoBEIR_mean
        metrics:
          - type: cosine_accuracy@1
            value: 0.4266666666666667
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.5533333333333333
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.6333333333333333
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.7333333333333334
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.4266666666666667
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.27111111111111114
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.22133333333333335
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.16266666666666665
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.21938427385320922
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.29853521291644186
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.3416857259806653
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.4179413679833615
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.42456607397934293
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.5169074074074074
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.33543474166245174
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: chowder val
          type: chowder-val
        metrics:
          - type: cosine_accuracy@1
            value: 0.952
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.9805
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.989
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.9955
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.952
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.32683333333333325
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.1978
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.09955000000000001
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.952
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.9805
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.989
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9955
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.9743050654346748
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.9674267857142853
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.9676369980648283
            name: Cosine Map@100

chowder-embed (SPECTER fine-tuned on the chowder corpus)

This is a sentence-transformers model finetuned from sentence-transformers/allenai-specter on the parquet dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: sentence-transformers/allenai-specter
Maximum Sequence Length: 512 tokens
Output Dimensionality: 768 dimensions
Similarity Function: Cosine Similarity
Supported Modality: Text
Training Dataset:
- parquet
Language: en

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 768, 'pooling_mode': 'cls', 'include_prompt': True})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

pip install -U sentence-transformers

Then you can load this model and run inference.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("Wakeleyhuh/chowder-embed-specter")
# Run inference
queries = [
    'ERBB3 blockade sensitizes hepatocellular carcinoma to regorafenib after first-line tyrosine kinase inhibitor resistance by inhibiting HIF1A-ABCB1 signaling',
]
documents = [
    'BACKGROUND/AIMS: Regorafenib is recommended by guidelines and trials as a sequential second-line therapy following progression on first-line sorafenib or lenvatinib in hepatocellular carcinoma (HCC). However, efficacy is limited, highlighting the urgent need to screen suitable patients and develop sensitization strategies. METHODS: Acquired sorafenib- or lenvatinib-resistant (SR or LR) HCC cell lines and organoids were established. Genome-wide CRISPR library screen was performed in SR or LR cell strains to identify synthetic lethal targets of regorafenib. RNA-seq and FITC-regorafenib efflux assay were used to elucidate ERBB3-driven downstream signaling. Preclinical mouse models of cell line- and patient-derived xenografts and clinical cohorts of HCC patients were employed to validate the efficacy of ERBB3-guided patient stratification. RESULTS: Screening with CRISPR library, we showed that inhibition of ERBB3 was synthetic lethal with regorafenib in SR or LR cell strains and organoids. Mechanistically, SR or LR triggered feedback activation of ERBB3 signaling and mediated regorafenib efflux via ERBB3-HIF1A-ABCB1 cascade pathway, limiting sensitivity to regorafenib. Moreover, ERBB3-low tumors following SR or LR exhibited significant sensitivity to regorafenib, suggesting its potential as a predictive biomarker to screen optimal candidates for sequential therapy. Seribantumab, an ERBB3-targeting monoclonal antibody, inhibited ERBB3-HIF1A-ABCB1 cascade, and its combination with regorafenib exerted marked synergistic anti-tumor effects on ERBB3-high tumors resistant to sorafenib or lenvatinib both in vitro and in vivo. CONCLUSIONS: This study revealed that ERBB3 was a key resistance factor driving limited efficacy to sequential regorafenib, but also an effective therapeutic target whose inhibition enhanced regorafenib sensitivity after SR or LR.',
    'We improve the theoretical predictions for rare $Z$-boson decays, $Z\\to V\\ell^+\\ell^-$ ($\\ell=e$ or $\\mu$), where $V$ denotes a heavy vector quarkonium including $J/\\Psi$, $\\Psi (2S)$, and $\\Upsilon (nS)$ with $n=1,2,3$. These processes are thought to be dominated by the electromagnetic fragmentation transition, i.e., $Z\\to \\gamma^*\\ell^+\\ell^-$ followed by $\\gamma^*\\rightarrow V$. The present study includes all of the relevant tree-level Feynman diagrams, which contribute to these decays in the standard model. Our analysis shows that, for the charmonium final states, the fragmentation transition almost saturates the whole contribution and the other diagrams can be neglected; while for the bottomonium final states, the inclusion of other diagrams can increase their branching fractions by $4\\%\\sim 9\\%$. Further investigation of the differential distributions, especially the angular distributions, indicates that forward-backward asymmetries for final leptons in these processes would be zero in the standard model. Therefore, in future experimental facilities with large number of $Z$-boson events accumulated, studies of these rare $Z$ decays may help both to test the standard model and to probe its interesting extensions.',
    'Historically, veterinary studies screening for breed, age and sex predisposition to disease have relied on collating small-scale studies of clinical datasets. The availability of larger datasets through groups such as the Small Animal Veterinary Surveillance Network (SAVSNET) promise access to information regarding a wide range of clinical presentations at scale, however, methodological limitations surrounding the extraction of specific disease information or screening for disease predispositions result in a substantial reduction in the number of animals studied. These studies often address very focused hypotheses - only leveraging a small fraction of the intrinsic value of the data at any one time. Here, we implemented an unsupervised machine learning methodology, creating a representation of a large volume of clinical notes collected by SAVSNET from veterinary practices across the UK. We utilise BERTopic, a topic-modelling tool based on Bidirectional Encoder Representations using Transformers (BERT) architecture, and show it is able to surface known phenotypes, such as breed predispositions to hypoadrenocorticism, diabetes mellitus and mitral valve disease, as well as potential novel patterns of disease phenotypes. This scalable and granular modelling technique facilitates the rapid interrogation of large clinical datasets, enabling the identification of a broad range of phenotypes within the population and the early detection of temporal changes indicative of emerging infectious or environmental diseases. Supplementary Information: The online version contains supplementary material available at 10.1186/s40537-026-01365-0.',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.8640, 0.0814, 0.0283]])

Evaluation

Metrics

Information Retrieval

Datasets: NanoSCIDOCS, NanoSciFact, NanoNFCorpus and chowder-val
Evaluated with InformationRetrievalEvaluator

Metric	NanoSCIDOCS	NanoSciFact	NanoNFCorpus	chowder-val
cosine_accuracy@1	0.38	0.6	0.3	0.952
cosine_accuracy@3	0.62	0.68	0.36	0.9805
cosine_accuracy@5	0.72	0.72	0.46	0.989
cosine_accuracy@10	0.86	0.8	0.54	0.9955
cosine_precision@1	0.38	0.6	0.3	0.952
cosine_precision@3	0.3133	0.2467	0.2533	0.3268
cosine_precision@5	0.256	0.16	0.248	0.1978
cosine_precision@10	0.176	0.09	0.222	0.0996
cosine_recall@1	0.0807	0.565	0.0125	0.952
cosine_recall@3	0.1947	0.67	0.0309	0.9805
cosine_recall@5	0.2627	0.715	0.0474	0.989
cosine_recall@10	0.3617	0.79	0.1022	0.9955
cosine_ndcg@10	0.3456	0.6831	0.245	0.9743
cosine_mrr@10	0.5344	0.6557	0.3606	0.9674
cosine_map@100	0.2648	0.6497	0.0919	0.9676

Nano BEIR

Dataset: NanoBEIR_mean

Evaluated with NanoBEIREvaluator with these parameters:

{
    "dataset_names": [
        "scidocs",
        "scifact",
        "nfcorpus"
    ],
    "dataset_id": "sentence-transformers/NanoBEIR-en"
}

Metric	Value
cosine_accuracy@1	0.4267
cosine_accuracy@3	0.5533
cosine_accuracy@5	0.6333
cosine_accuracy@10	0.7333
cosine_precision@1	0.4267
cosine_precision@3	0.2711
cosine_precision@5	0.2213
cosine_precision@10	0.1627
cosine_recall@1	0.2194
cosine_recall@3	0.2985
cosine_recall@5	0.3417
cosine_recall@10	0.4179
cosine_ndcg@10	0.4246
cosine_mrr@10	0.5169
cosine_map@100	0.3354

Training Details

Training Dataset

parquet

Dataset: parquet
Size: 70,764 training samples
Columns: anchor and positive
Approximate statistics based on the first 100 samples:
anchor positive
type string string

modality text text
details
min: 8 tokens
mean: 18.96 tokens
max: 51 tokens

min: 54 tokens
mean: 276.39 tokens
max: 512 tokens

	anchor	positive
type	string	string
modality	text	text
details	min: 8 tokens mean: 18.96 tokens max: 51 tokens	min: 54 tokens mean: 276.39 tokens max: 512 tokens

Samples:

anchor	positive
`Los-mamba: A low-rank recursive mamba framework for mitigating stationary bias in trajectory prediction`	Trajectory prediction plays a key role in autonomous driving and intelligent transportation systems. Mamba performs well in modeling long sequences but struggles with short-term static or local motion features. In this paper, we propose the Lo w-rank Recur s ive Mamba framework ( Los-Mamba ). It addresses prediction bias caused by inadequate modeling of local static features in trajectory prediction tasks. Firstly, Los-Mamba uses hybrid Mamba branches to capture both global motion trends and local features. Secondly, it reduces trajectory feature dimensionality through low-rank approximation, simplifying computational complexity. Thirdly, Los-Mamba also includes a spatio-temporal constraint mechanism. It uses position coding and a multilayer perceptron (MLP) to interpolate static and dynamic trajectories, improving the representation of local features. Experiments show that Los-Mamba outperforms baseline models on datasets like ETH and Hotel. Its average displacement error (ADE) and fi...
`The Universal Language of CSI:Unifying Wireless Sensing Across Devices and Environments`	WiFi sensing based on Channel State Information (CSI) promises ubiquitous, device-free perception, yet current research remains trapped in a Tower of Babel - fragmented into isolated silos where models are tailored to specific hardware dialects, fixed environments, and narrow tasks. The primary bottleneck is the Heterogeneity Gap: the disparity in signal dimensions, sampling rates, and semantic labels that prevents cross-system understanding. To bridge this gap, we propose a foundation-model framework that treats CSI not merely as raw signals but as a structured language with a learnable universal grammar. We first curate and standardize a large collection of heterogeneous real-world CSI datasets, establishing a unified infrastructure that allows incompatible signal formats to be treated as a single corpus. Second, we introduce a modular architecture that acts as a universal translator where lightweight dataset-specific adapters tokenize diverse signal inputs into a shared latent vocab...
`Long-term follow-up of the transplant-eligible cohort of the EMN12/HOVON-129 study for primary plasma cell leukemia patients`	Primary plasma cell leukemia (pPCL) is a rare and aggressive plasma cell disorder [ 1 , 2 ]. Compared to multiple myeloma (MM), pPCL is characterized by a higher frequency of poor-risk features, such as high-risk cytogenetic abnormalities (HRCAs) and elevated LDH [ 1 , 2 ]. Although MM treatment has improved, the survival of pPCL patients remains poor [ 1 , 2 ]. We therefore initiated the phase 2 EMN12/HOVON-129 study for patients with newly diagnosed pPCL, which showed that carfilzomib and lenalidomide-based therapy in conjunction with autologous stem-cell-transplantation (ASCT) provides efficient disease control [ 3 ]. Here, we present the long-term follow-up of patients aged 18–65 years included in the EMN12/HOVON-129 study. We also aimed to evaluate if patients with HRCAs, elevated LDH, or other high-risk features have a worse clinical outcome. Inclusion criteria were newly diagnosed pPCL (defined as >2 × 10 9 /L circulating plasma cells or >20% plasma cells in peripheral blood) an...

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}

Evaluation Dataset

parquet

Dataset: parquet
Size: 2,000 evaluation samples
Columns: anchor and positive
Approximate statistics based on the first 100 samples:
anchor positive
type string string

modality text text
details
min: 7 tokens
mean: 19.21 tokens
max: 37 tokens

min: 68 tokens
mean: 278.03 tokens
max: 512 tokens

	anchor	positive
type	string	string
modality	text	text
details	min: 7 tokens mean: 19.21 tokens max: 37 tokens	min: 68 tokens mean: 278.03 tokens max: 512 tokens

Samples:

anchor	positive
`Boolean Algebra -- Driven Sepsis Diagnosis`	Sepsis remains a diagnostic challenge due to its heterogeneous molecular signatures and complex immune responses. In this study, we develop a logical data analysis framework based on Boolean polynomial rings. This method constructs an ideal $\mathcal{I}$ of selection criteria that isolate empty subsets of previously analyzed patient data. This approach enables the derivation of interpretable classification rules based on biomarker profiles. We demonstrate that logical data analysis identifies distinct logical patterns for positive and negative sepsis classification. For instance, elevated levels of GLP-1 and MyD88 are associated with septic states in our dataset, whereas high TRAIL and low MyD88 concentrations may suggest a non-septic condition. Importantly, a new way to integrate expert knowledge to filter out potential overfitting or dataset-specific artifacts is shown. Our findings highlight the utility of logics in generating transparent, biologically plausible rules for a data-bas...
`Notes on the bootstrap of four-point conformal integrals`	We set up a bootstrap workflow to study four-point conformal integrals in position space, using leading singularities, single-valued multiple polylogarithmic ans"atze and boundary data from expansion by regions. These four-point conformal integrals are general in the sense that they are generated by the four-point projections of all possible $f$-graphs, including all non-planar $f$-graph sectors. For three-loop cases, fourteen of the fifteen inequivalent integrand basis can be directly calculated by \texttt{HyperlogProcedures} and the last one is fixed by Gram identity. Then we concentrate on how far the bootstrap workflow can go for four-loop cases, though it works for three-loop cases as well. We show that integrals with several leading singularities can be made tractable by decomposing them into pieces with simpler cut structure. Some four-loop integrals which can not be calculated or very hard to be calculated by other methods for now are obtained in this way. We also provide a pa...
`Clean2FX: Label-conditioned modeling for clean-to-effect guitar audio transformations`	We present Clean2FX, a study and demo of label-conditioned clean-to-effect transformation for electric guitar audio. Given a clean guitar input and a target effect label, the task is to synthesize the corresponding effected signal while preserving the musical content. Training and evaluation pairs are constructed from EGFxSet real, single tone recordings by assembling matched clean/effected chords, melodies, and mixed timelines. This allows for controlled comparison across effects. We evaluate four neural approaches under a common spectrogram-based transformation setting: two variational autoencoders and two U-Net models that differ in whether they operate on linear or log-magnitude representations. Performance is measured using linear-magnitude spectrogram MSE and Fr'echet Audio Distance. The U-Net models outperform the variational autoencoder variants. Per-effect results show that distortion effects are most readily improved, whereas delay and reverb effects exhibit weaker FAD gains...

Loss: MultipleNegativesRankingLoss with these parameters:

{
    "scale": 20.0,
    "similarity_fct": "cos_sim",
    "gather_across_devices": false,
    "directions": [
        "query_to_doc"
    ],
    "partition_mode": "joint",
    "hardness_mode": null,
    "hardness_strength": 0.0
}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 16
learning_rate: 2e-05
warmup_steps: 0.1
weight_decay: 0.01
fp16: True
per_device_eval_batch_size: 16
load_best_model_at_end: True
seed: 12
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

per_device_train_batch_size: 16
num_train_epochs: 3
max_steps: -1
learning_rate: 2e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0.1
optim: adamw_torch_fused
optim_args: None
weight_decay: 0.01
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
optim_target_modules: None
gradient_accumulation_steps: 1
average_tokens_across_devices: True
max_grad_norm: 1.0
label_smoothing_factor: 0.0
bf16: False
fp16: True
bf16_full_eval: False
fp16_full_eval: False
tf32: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
use_liger_kernel: False
liger_kernel_config: None
use_cache: False
neftune_noise_alpha: None
torch_empty_cache_steps: None
auto_find_batch_size: False
log_on_each_node: True
logging_nan_inf_filter: True
include_num_input_tokens_seen: no
log_level: passive
log_level_replica: warning
disable_tqdm: False
project: huggingface
trackio_space_id: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 16
prediction_loss_only: True
eval_on_start: False
eval_do_concat_batches: True
eval_use_gather_object: False
eval_accumulation_steps: None
include_for_metrics: []
batch_eval_metrics: False
save_only_model: False
save_on_each_node: False
enable_jit_checkpoint: False
push_to_hub: False
hub_private_repo: None
hub_model_id: None
hub_strategy: every_save
hub_always_push: False
hub_revision: None
load_best_model_at_end: True
ignore_data_skip: False
restore_callback_states_from_checkpoint: False
full_determinism: False
seed: 12
data_seed: None
use_cpu: False
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
parallelism_config: None
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_pin_memory: True
dataloader_persistent_workers: False
dataloader_prefetch_factor: None
dataloader_multiprocessing_context: None
dataloader_in_order: True
remove_unused_columns: True
label_names: None
train_sampling_strategy: random
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
ddp_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
local_rank: -1
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}
warmup_ratio: None

Training Logs

Epoch	Step	Training Loss	Validation Loss	NanoSCIDOCS_cosine_ndcg@10	NanoSciFact_cosine_ndcg@10	NanoNFCorpus_cosine_ndcg@10	NanoBEIR_mean_cosine_ndcg@10	chowder-val_cosine_ndcg@10
-1	-1	-	-	0.3211	0.6251	0.1796	0.3753	0.8622
0.0002	1	0.1790	-	-	-	-	-	-
0.0301	133	0.1970	-	-	-	-	-	-
0.0601	266	0.0910	-	-	-	-	-	-
0.0902	399	0.0571	-	-	-	-	-	-
0.1203	532	0.0445	-	-	-	-	-	-
0.1504	665	0.0331	-	-	-	-	-	-
0.1804	798	0.0311	-	-	-	-	-	-
0.2105	931	0.0304	-	-	-	-	-	-
0.2406	1064	0.0234	-	-	-	-	-	-
0.2706	1197	0.0229	-	-	-	-	-	-
0.3000	1327	-	0.0102	0.3527	0.6570	0.2448	0.4182	-
0.3007	1330	0.0225	-	-	-	-	-	-
0.3308	1463	0.0199	-	-	-	-	-	-
0.3608	1596	0.0220	-	-	-	-	-	-
0.3909	1729	0.0143	-	-	-	-	-	-
0.4210	1862	0.0186	-	-	-	-	-	-
0.4511	1995	0.0136	-	-	-	-	-	-
0.4811	2128	0.0191	-	-	-	-	-	-
0.5112	2261	0.0151	-	-	-	-	-	-
0.5413	2394	0.0220	-	-	-	-	-	-
0.5713	2527	0.0170	-	-	-	-	-	-
0.6	2654	-	0.0069	0.3455	0.6831	0.245	0.4245	-
0.6014	2660	0.0150	-	-	-	-	-	-
0.6315	2793	0.0174	-	-	-	-	-	-
0.6615	2926	0.0123	-	-	-	-	-	-
0.6916	3059	0.0178	-	-	-	-	-	-
0.7217	3192	0.0182	-	-	-	-	-	-
0.7518	3325	0.0151	-	-	-	-	-	-
0.7818	3458	0.0143	-	-	-	-	-	-
0.8119	3591	0.0124	-	-	-	-	-	-
0.8420	3724	0.0165	-	-	-	-	-	-
0.8720	3857	0.0134	-	-	-	-	-	-
0.9001	3981	-	0.0047	0.3545	0.6515	0.2247	0.4102	-
0.9021	3990	0.0119	-	-	-	-	-	-
0.9322	4123	0.0156	-	-	-	-	-	-
0.9622	4256	0.0127	-	-	-	-	-	-
0.9923	4389	0.0154	-	-	-	-	-	-
1.0224	4522	0.0091	-	-	-	-	-	-
1.0525	4655	0.0034	-	-	-	-	-	-
1.0825	4788	0.0065	-	-	-	-	-	-
1.1126	4921	0.0072	-	-	-	-	-	-
1.1427	5054	0.0041	-	-	-	-	-	-
1.1727	5187	0.0065	-	-	-	-	-	-
1.2001	5308	-	0.0032	0.3672	0.6294	0.2201	0.4055	-
1.2028	5320	0.0055	-	-	-	-	-	-
1.2329	5453	0.0038	-	-	-	-	-	-
1.2629	5586	0.0043	-	-	-	-	-	-
1.2930	5719	0.0040	-	-	-	-	-	-
1.3231	5852	0.0038	-	-	-	-	-	-
1.3532	5985	0.0069	-	-	-	-	-	-
1.3832	6118	0.0063	-	-	-	-	-	-
1.4133	6251	0.0052	-	-	-	-	-	-
1.4434	6384	0.0068	-	-	-	-	-	-
1.4734	6517	0.0047	-	-	-	-	-	-
1.5001	6635	-	0.0032	0.3589	0.6404	0.1966	0.3986	-
-1	-1	-	-	0.3456	0.6831	0.2450	0.4246	0.9743

The bold row denotes the saved checkpoint.

Training Time

Training: 44.8 minutes
Evaluation: 3.9 minutes
Total: 48.7 minutes

Framework Versions

Python: 3.14.7
Sentence Transformers: 5.7.0
Transformers: 5.15.0
PyTorch: 2.13.0+cu130
Accelerate: 1.14.0
Datasets: 5.0.1
Tokenizers: 0.22.2

Additional Resources

Training and Finetuning Embedding Models with Sentence Transformers: the end-to-end guide for training or finetuning Sentence Transformer models.
Introduction to Matryoshka Embedding Models: variable-size embeddings that can be truncated with minimal quality loss.
Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval: post-training compression of embedding vectors.
Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video models through the same API.
Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: train multimodal embedding models, with a Visual Document Retrieval walkthrough.

Citation

BibTeX

Sentence Transformers

@inproceedings{reimers-2019-sentence-bert,
    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
    author = "Reimers, Nils and Gurevych, Iryna",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
    month = "11",
    year = "2019",
    publisher = "Association for Computational Linguistics",
    url = "https://arxiv.org/abs/1908.10084",
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}