quantosimo-curriculum-embeddings-v1

QuantOsimo Curriculum Embeddings v1

A fine-tuned sentence-transformer embedding model for UK curriculum content retrieval.

Model Details

• Base model: sentence-transformers/all-MiniLM-L6-v2
• Fine-tuned on: 1,077 curriculum triplets from 1,714 official AQA/Edexcel/OCR specification chunks (KS2-KS5, Maths and Science)
• Training: MultipleNegativesRankingLoss, 3 epochs, batch size 4
• Embedding dimension: 384 (fits pgvector(384))

Performance

• Baseline triplet accuracy: 0.9252
• Fine-tuned triplet accuracy: 0.9907
• Error rate reduction: 88%

Intended Use

Semantic retrieval of UK curriculum content for QuantOsimo (quantosimo.com) — an AI-powered Maths and Science tutoring platform for UK students KS2 to A-Level.

Training Data

Generated from official AQA, Edexcel, and OCR specification documents across KS2-KS5 Maths and Science.

Built by Samuel Osagiede as part of QuantOsimo's RAG pipeline.

---

SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: sentence-transformers/all-MiniLM-L6-v2 <!-- at revision 1110a243fdf4706b3f48f1d95db1a4f5529b4d41 -->
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text

<!-- - **Training Dataset:** Unknown --> <!-- - **Language:** Unknown --> <!-- - **License:** Unknown -->

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': 384, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Normalize({})
)

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("sentence_transformers_model_id")
# Run inference
sentences = [
    'Physics. Students study energy, including energy stores and transfers, calculations of kinetic, gravitational potential and elastic potential energy, power, efficiency and the national and global use of energy resources.',
    'Students study energy, including energy stores and transfers, calculations of kinetic, gravitational potential and elastic potential energy, power, efficiency and the national and global use of energy resources. Electricity covers circuit diagrams, current, potential difference and resistance, series and parallel circuits, the I–V characteristics of components, mains electricity, direct and alternating current, the national grid, and electrical power and energy calculations.',
    "Pearson Edexcel Level 3 Advanced GCE in Biology B (9BI0) First teaching from September 2015 First certification from 2017 Level 3 Advanced GCE First certification 2017 Edexcel, BTEC and LCCI qualifications Edexcel, BTEC and LCCI qualifications are awarded by Pearson, the UK’s largest awarding body offering academic and vocational qualifications that are globally recognised and benchmarked. For further information, please visit our qualifications website at qualifications.pearson.com. Alternatively, you can get in touch with us using the details on our contact us page at qualifications.pearson.com/contactus Pearson is the world's leading learning company, with 35,000 employees in more than 70 countries working to help people of all ages to make measurable progress in their lives through learning. We put the learner at the centre of everything we do, because wherever learning flourishes, so do people. Find out more about how we can help you and your learners at qualifications.pearson.com References to third party material made in this specification are made in good faith. Pearson does not endorse, approve or accept responsibility for the content of materials, which may be subject to change, or any opinions expressed therein. (Material may include textbooks, journals, magazines and other publications and All information in this specification is correct at time of publication. Original origami artwork: Mark Bolitho Origami photography: Pearson Education Ltd/Naki Kouyioumtzis ISBN 978 1 446 93798 3 All the material in this publication is copyright © Pearson Education Limited 2018 From Pearson’s Expert Panel for World Class Qualifications The reform of the qualifications system in England is a profoundly important change to the education system. Teachers need to know that the new qualifications will assist them in helping their learners make progress in their lives. When these changes were first proposed we were approached by Pearson to join an ‘Expert Panel’ that would advise them on the development of the new qualifications. We were chosen, either because of our expertise in the UK education system, or because of our experience in reforming qualifications in other systems around the world as diverse as Singapore, Hong Kong, Australia and a number of countries We have guided Pearson through what we judge to be a rigorous qualification development process that has included: extensive international comparability of subject content against the highestperforming jurisdictions in the world benchmarking assessments against UK and overseas providers to ensure that they are at the right level of demand establishing External Subject Advisory Groups, drawing on independent subjectspecific expertise to challenge and validate our qualifications subjecting the final qualifications to scrutiny against the DfE content and Ofqual accreditation criteria in advance of submission. Importantly, we have worked to ensure that the content and learning is future oriented. The design has been guided by what is called an ‘Efficacy Framework’, meaning learner outcomes have been at the heart of this development throughout. We understand that ultimately it is excellent teaching that is the key factor to a learner’s success in education. As a result of our work as a panel we are confident that",
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000,  0.6863, -0.0800],
#         [ 0.6863,  1.0000, -0.0040],
#         [-0.0800, -0.0040,  1.0000]])

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* -->

Evaluation

Metrics

Triplet

• Dataset: curriculum-eval
• Evaluated with <code>TripletEvaluator</code>

<!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Dataset

Unnamed Dataset

• Size: 970 training samples
• Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>sentence_2</code>
• Approximate statistics based on the first 100 samples:

  	sentence_0	sentence_1	sentence_2
  type	string	string	string
  modality	text	text	text
  details	<ul><li>min: 15 tokens</li><li>mean: 194.26 tokens</li><li>max: 256 tokens</li></ul>	<ul><li>min: 38 tokens</li><li>mean: 196.3 tokens</li><li>max: 256 tokens</li></ul>	<ul><li>min: 16 tokens</li><li>mean: 177.9 tokens</li><li>max: 256 tokens</li></ul>

• Samples:

  sentence_0	sentence_1	sentence_2
  <code>evidence for analysis. They can see how accurate and sensitive methods of analysis can be applied to the study of chemical changes, but also to detect drugs such as in blood or urine testing in sport. Topic 7A: Mass spectrometry 1. be able to use data from a mass spectrometer to: determine the relative molecular mass of an organic compound from the ii suggest possible structures of a simple organic compound from the m/z of the molecular ion and fragmentation patterns Topic 7B: Infrared (IR) spectroscopy 2. be able to use data from infrared spectra to deduce functional groups present in organic compounds and to predict infrared absorptions, given wavenumber data, due to familiar functional groups, including: C–H stretching absorption in alkanes, alkenes and aldehydes ii C=C stretching absorption in alkenes iii O–H stretching absorption in alcohols iv C=O stretching absorption in aldehydes and ketones v C=O stretching absorption and the broad O-H stretching absorption in vi N–H stretchin...</code>	<code>of: 5. understand experiments to measure enthalpy changes in terms of: processing results using the expression: energy transferred = mass x specific heat capacity × temperature change ii evaluating sources of error and assumptions made in the experiments Students will need to consider experiments where: substances are mixed in an insulated container and the temperature enthalpy of combustion is measured, such as using a series of alcohols in a the enthalpy change cannot be measured directly. 6. be able to calculate enthalpy changes in kJ mol-1 from given experimental results Both a sign and units are expected in the final answer. 7. be able to construct enthalpy cycles using Hess’s Law 8. be able to calculate enthalpy changes from data using Hess’s Law CORE PRACTICAL 8: To determine the enthalpy change of a reaction using 9. know what is meant by the terms ‘bond enthalpy’ and ‘mean bond enthalpy’ 10. be able to calculate an enthalpy change of reaction using mean bond enthalpies and exp...</code>	<code>A-Level Mathematics (AQA 7357) — Pure Mathematics.</code>
  <code>system. Explain the role of meiotic cell division, including the production of four daughter cells, each with half the number of chromosomes, and that this results in the formation of genetically different haploid gametes The stages of meiosis are not required Describe DNA as a polymer made up of: a two strands coiled to form a double helix b strands linked by a series of complementary base pairs joined together by weak hydrogen bonds nucleotides that consist of a sugar and phosphate group with one of the four different bases attached to the sugar Describe the genome as the entire DNA of an organism and a gene as a section of a DNA molecule that codes for a specific Explain how DNA can be extracted from fruit 3.12 Explain why there are differences in the inherited characteristics as a result of alleles 3.13 Explain the terms: chromosome, gene, allele, dominant, recessive, homozygous, heterozygous, genotype, phenotype, 3.14 Explain monohybrid inheritance using genetic diagrams, Punnett ...</code>	<code>arithmetic means (2b). Suggested practicals ● Investigate the variations in a species to illustrate continuous variation and discontinuous ● Investigate inheritance using suitable organisms or models. Topic 4 – Natural selection and genetic modification Explain Charles Darwin’s theory of evolution by natural selection Explain how the emergence of resistant organisms supports Charles Darwin’s theory of evolution including antibiotic resistance in bacteria Describe the evidence for human evolution, based on fossils, a Ardi from 4.4 million years ago b Lucy from 3.2 million years ago Richard Leakey’s discovery of fossils from 1.6 million Describe the evidence for human evolution based on stone a the development of stone tools over time b how these can be dated from their environment Describe how genetic analysis has led to the suggestion of the three domains rather than the five kingdoms classification Explain selective breeding and its impact on food plants and domesticated animals 4.10 ...</code>	<code>Pearson Edexcel Level 3 Advanced GCE in Biology B (9BI0) First teaching from September 2015 First certification from 2017 Level 3 Advanced GCE First certification 2017 Edexcel, BTEC and LCCI qualifications Edexcel, BTEC and LCCI qualifications are awarded by Pearson, the UK’s largest awarding body offering academic and vocational qualifications that are globally recognised and benchmarked. For further information, please visit our qualifications website at qualifications.pearson.com. Alternatively, you can get in touch with us using the details on our contact us page at qualifications.pearson.com/contactus Pearson is the world's leading learning company, with 35,000 employees in more than 70 countries working to help people of all ages to make measurable progress in their lives through learning. We put the learner at the centre of everything we do, because wherever learning flourishes, so do people. Find out more about how we can help you and your learners at qualifications.pearson.com...</code>
  <code>Ratio, proportion and rates of change. Pupils change freely between related standard units (for example time, length, area, volume and mass) and use ratio notation, reducing a ratio to its simplest form.</code>	<code>Pupils change freely between related standard units (for example time, length, area, volume and mass) and use ratio notation, reducing a ratio to its simplest form. They divide a given quantity into two or more parts in a given part-to-part or part-to-whole ratio, express a multiplicative relationship between two quantities as a ratio or a fraction, and understand and use proportion as equality of ratios.</code>	<code>Key Stage 3 Science — Biology: Cells and Organisation. Pupils should be taught that cells are the fundamental unit of living organisms, including how to observe, interpret and record cell structure using a light microscope.</code>

• Loss: <code>MultipleNegativesRankingLoss</code> 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: 4
• per_device_eval_batch_size: 4
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

<details><summary>Click to expand</summary>

• per_device_train_batch_size: 4
• num_train_epochs: 3
• max_steps: -1
• learning_rate: 5e-05
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 0
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.0
• 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
• label_smoothing_factor: 0.0
• bf16: False
• fp16: False
• 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: 4
• 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: False
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• 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
• 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
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

</details>

Training Logs

Training Time

• Training: 28.4 minutes

Framework Versions

• Python: 3.12.1
• Sentence Transformers: 5.6.0
• Transformers: 5.13.1
• PyTorch: 2.13.0+cu130
• Accelerate: 1.14.0
• Datasets: 5.0.0
• Tokenizers: 0.22.2

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},
}

<!-- ## Glossary *Clearly define terms in order to be accessible across audiences.* --> <!-- ## Model Card Authors *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.* --> <!-- ## Model Card Contact *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.* -->