metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:156
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
base_model: Snowflake/snowflake-arctic-embed-l
widget:
  - source_sentence: When did Google release their gemini-20-flash-thinking-exp model?
    sentences:
      - >-
        Your browser does not support the audio element.


        OpenAI aren’t the only group with a multi-modal audio model. Google’s
        Gemini also accepts audio input, and the Google Gemini apps can speak in
        a similar way to ChatGPT now. Amazon also pre-announced voice mode for
        Amazon Nova, but that’s meant to roll out in Q1 of 2025.

        Google’s NotebookLM, released in September, took audio output to a new
        level by producing spookily realistic conversations between two “podcast
        hosts” about anything you fed into their tool. They later added custom
        instructions, so naturally I turned them into pelicans:



        Your browser does not support the audio element.
      - >-
        OpenAI are not the only game in town here. Google released their first
        entrant in the category, gemini-2.0-flash-thinking-exp, on December
        19th.

        Alibaba’s Qwen team released their QwQ model on November 28th—under an
        Apache 2.0 license, and that one I could run on my own machine. They
        followed that up with a vision reasoning model called QvQ on December
        24th, which I also ran locally.

        DeepSeek made their DeepSeek-R1-Lite-Preview model available to try out
        through their chat interface on November 20th.

        To understand more about inference scaling I recommend Is AI progress
        slowing down? by Arvind Narayanan and Sayash Kapoor.
      - >-
        The earliest of those was Google’s Gemini 1.5 Pro, released in February.
        In addition to producing GPT-4 level outputs, it introduced several
        brand new capabilities to the field—most notably its 1 million (and then
        later 2 million) token input context length, and the ability to input
        video.

        I wrote about this at the time in The killer app of Gemini Pro 1.5 is
        video, which earned me a short appearance as a talking head in the
        Google I/O opening keynote in May.
  - source_sentence: >-
      Which organizations currently have models that score higher than
      GPT-4-0314 on the Chatbot Arena leaderboard?
    sentences:
      - |-
        blogging
                    105


                    ai
                    1250


                    generative-ai
                    1077


                    llms
                    1065

        Next: Tom Scott, and the formidable power of escalating streaks
        Previous: Last weeknotes of 2023


         
         


        Colophon
        ©
        2002
        2003
        2004
        2005
        2006
        2007
        2008
        2009
        2010
        2011
        2012
        2013
        2014
        2015
        2016
        2017
        2018
        2019
        2020
        2021
        2022
        2023
        2024
        2025
      - >-
        Then there’s the rest. If you browse the Chatbot Arena leaderboard
        today—still the most useful single place to get a vibes-based evaluation
        of models—you’ll see that GPT-4-0314 has fallen to around 70th place.
        The 18 organizations with higher scoring models are Google, OpenAI,
        Alibaba, Anthropic, Meta, Reka AI, 01 AI, Amazon, Cohere, DeepSeek,
        Nvidia, Mistral, NexusFlow, Zhipu AI, xAI, AI21 Labs, Princeton and
        Tencent.

        Training a GPT-4 beating model was a huge deal in 2023. In 2024 it’s an
        achievement that isn’t even particularly notable, though I personally
        still celebrate any time a new organization joins that list.

        Some of those GPT-4 models run on my laptop
      - >-
        If you think about what they do, this isn’t such a big surprise. The
        grammar rules of programming languages like Python and JavaScript are
        massively less complicated than the grammar of Chinese, Spanish or
        English.

        It’s still astonishing to me how effective they are though.

        One of the great weaknesses of LLMs is their tendency to hallucinate—to
        imagine things that don’t correspond to reality. You would expect this
        to be a particularly bad problem for code—if an LLM hallucinates a
        method that doesn’t exist, the code should be useless.
  - source_sentence: >-
      Why is gullibility considered the biggest unsolved problem in the context
      of large language models?
    sentences:
      - >-
        I think this means that, as individual users, we don’t need to feel any
        guilt at all for the energy consumed by the vast majority of our
        prompts. The impact is likely neglible compared to driving a car down
        the street or maybe even watching a video on YouTube.

        Likewise, training. DeepSeek v3 training for less than $6m is a
        fantastic sign that training costs can and should continue to drop.

        For less efficient models I find it useful to compare their energy usage
        to commercial flights. The largest Llama 3 model cost about the same as
        a single digit number of fully loaded passenger flights from New York to
        London. That’s certainly not nothing, but once trained that model can be
        used by millions of people at no extra training cost.
      - |-
        Large Language Models
        They’re actually quite easy to build
        You can run LLMs on your own devices
        Hobbyists can build their own fine-tuned models
        We don’t yet know how to build GPT-4
        Vibes Based Development
        LLMs are really smart, and also really, really dumb
        Gullibility is the biggest unsolved problem
        Code may be the best application
        The ethics of this space remain diabolically complex
        My blog in 2023
      - >-
        One way to think about these models is an extension of the
        chain-of-thought prompting trick, first explored in the May 2022 paper
        Large Language Models are Zero-Shot Reasoners.

        This is that trick where, if you get a model to talk out loud about a
        problem it’s solving, you often get a result which the model would not
        have achieved otherwise.

        o1 takes this process and further bakes it into the model itself. The
        details are somewhat obfuscated: o1 models spend “reasoning tokens”
        thinking through the problem that are not directly visible to the user
        (though the ChatGPT UI shows a summary of them), then outputs a final
        result.
  - source_sentence: >-
      Which other AI-related articles were posted around the same time as the
      "Industry’s Tardy Response" article according to the provided context?
    sentences:
      - >-
        We already knew LLMs were spookily good at writing code. If you prompt
        them right, it turns out they can build you a full interactive
        application using HTML, CSS and JavaScript (and tools like React if you
        wire up some extra supporting build mechanisms)—often in a single
        prompt.

        Anthropic kicked this idea into high gear when they released Claude
        Artifacts, a groundbreaking new feature that was initially slightly lost
        in the noise due to being described half way through their announcement
        of the incredible Claude 3.5 Sonnet.

        With Artifacts, Claude can write you an on-demand interactive
        application and then let you use it directly inside the Claude
        interface.

        Here’s my Extract URLs app, entirely generated by Claude:
      - >-
        I think this means that, as individual users, we don’t need to feel any
        guilt at all for the energy consumed by the vast majority of our
        prompts. The impact is likely neglible compared to driving a car down
        the street or maybe even watching a video on YouTube.

        Likewise, training. DeepSeek v3 training for less than $6m is a
        fantastic sign that training costs can and should continue to drop.

        For less efficient models I find it useful to compare their energy usage
        to commercial flights. The largest Llama 3 model cost about the same as
        a single digit number of fully loaded passenger flights from New York to
        London. That’s certainly not nothing, but once trained that model can be
        used by millions of people at no extra training cost.
      - >-
        Industry’s Tardy Response to the AI Prompt Injection Vulnerability on
        RedMonk Conversations



        Posted 31st December 2023 at 11:59 pm · Follow me on Mastodon, Bluesky,
        Twitter or subscribe to my newsletter



        More recent articles


        Qwen 3 offers a case study in how to effectively release a model - 29th
        April 2025

        Watching o3 guess a photo's location is surreal, dystopian and wildly
        entertaining - 26th April 2025

        Exploring Promptfoo via Dave Guarino's SNAP evals - 24th April 2025


         


        This is Stuff we figured out about AI in 2023 by Simon Willison, posted
        on 31st December 2023.


        Part of series LLMs annual review


        Stuff we figured out about AI in 2023 - Dec. 31, 2023, 11:59 p.m. 

        Things we learned about LLMs in 2024 - Dec. 31, 2024, 6:07 p.m.
  - source_sentence: >-
      Why does the author find it astonishing that models like GPT-4 can run on
      their current hardware?
    sentences:
      - >-
        This remains astonishing to me. I thought a model with the capabilities
        and output quality of GPT-4 needed a datacenter class server with one or
        more $40,000+ GPUs.

        These models take up enough of my 64GB of RAM that I don’t run them
        often—they don’t leave much room for anything else.

        The fact that they run at all is a testament to the incredible training
        and inference performance gains that we’ve figured out over the past
        year. It turns out there was a lot of low-hanging fruit to be harvested
        in terms of model efficiency. I expect there’s still more to come.
      - >-
        I run a bunch of them on my laptop. I run Mistral 7B (a surprisingly
        great model) on my iPhone. You can install several different apps to get
        your own, local, completely private LLM. My own LLM project provides a
        CLI tool for running an array of different models via plugins.

        You can even run them entirely in your browser using WebAssembly and the
        latest Chrome!

        Hobbyists can build their own fine-tuned models

        I said earlier that building an LLM was still out of reach of hobbyists.
        That may be true for training from scratch, but fine-tuning one of those
        models is another matter entirely.
      - >-
        Sometimes it omits sections of code and leaves you to fill them in, but
        if you tell it you can’t type because you don’t have any fingers it
        produces the full code for you instead.

        There are so many more examples like this. Offer it cash tips for better
        answers. Tell it your career depends on it. Give it positive
        reinforcement. It’s all so dumb, but it works!

        Gullibility is the biggest unsolved problem

        I coined the term prompt injection in September last year.

        15 months later, I regret to say that we’re still no closer to a robust,
        dependable solution to this problem.

        I’ve written a ton about this already.

        Beyond that specific class of security vulnerabilities, I’ve started
        seeing this as a wider problem of gullibility.
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: SentenceTransformer based on Snowflake/snowflake-arctic-embed-l
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy@1
            value: 0.9166666666666666
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 1
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 1
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 1
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.9166666666666666
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.3333333333333333
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.20000000000000004
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.10000000000000002
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.9166666666666666
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 1
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 1
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 1
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.9637887397321441
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.9513888888888888
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.9513888888888888
            name: Cosine Map@100

SentenceTransformer based on Snowflake/snowflake-arctic-embed-l

This is a sentence-transformers model finetuned from Snowflake/snowflake-arctic-embed-l. It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: Snowflake/snowflake-arctic-embed-l
Maximum Sequence Length: 512 tokens
Output Dimensionality: 1024 dimensions
Similarity Function: Cosine Similarity

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, '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("szybe/legal-ft-b7a94f9f-5de7-476a-baa7-a2e80c3f10c1")
# Run inference
sentences = [
    'Why does the author find it astonishing that models like GPT-4 can run on their current hardware?',
    'This remains astonishing to me. I thought a model with the capabilities and output quality of GPT-4 needed a datacenter class server with one or more $40,000+ GPUs.\nThese models take up enough of my 64GB of RAM that I don’t run them often—they don’t leave much room for anything else.\nThe fact that they run at all is a testament to the incredible training and inference performance gains that we’ve figured out over the past year. It turns out there was a lot of low-hanging fruit to be harvested in terms of model efficiency. I expect there’s still more to come.',
    'Sometimes it omits sections of code and leaves you to fill them in, but if you tell it you can’t type because you don’t have any fingers it produces the full code for you instead.\nThere are so many more examples like this. Offer it cash tips for better answers. Tell it your career depends on it. Give it positive reinforcement. It’s all so dumb, but it works!\nGullibility is the biggest unsolved problem\nI coined the term prompt injection in September last year.\n15 months later, I regret to say that we’re still no closer to a robust, dependable solution to this problem.\nI’ve written a ton about this already.\nBeyond that specific class of security vulnerabilities, I’ve started seeing this as a wider problem of gullibility.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9167
cosine_accuracy@3	1.0
cosine_accuracy@5	1.0
cosine_accuracy@10	1.0
cosine_precision@1	0.9167
cosine_precision@3	0.3333
cosine_precision@5	0.2
cosine_precision@10	0.1
cosine_recall@1	0.9167
cosine_recall@3	1.0
cosine_recall@5	1.0
cosine_recall@10	1.0
cosine_ndcg@10	0.9638
cosine_mrr@10	0.9514
cosine_map@100	0.9514

Training Details

Training Dataset

Unnamed Dataset

Size: 156 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 156 samples:
sentence_0 sentence_1
type string string
details
min: 12 tokens
mean: 21.03 tokens
max: 34 tokens

min: 43 tokens
mean: 135.28 tokens
max: 214 tokens

	sentence_0	sentence_1
type	string	string
details	min: 12 tokens mean: 21.03 tokens max: 34 tokens	min: 43 tokens mean: 135.28 tokens max: 214 tokens

Samples:

sentence_0	sentence_1
`What are the new capabilities introduced by Google’s Gemini 15 Pro?`	`The earliest of those was Google’s Gemini 1.5 Pro, released in February. In addition to producing GPT-4 level outputs, it introduced several brand new capabilities to the field—most notably its 1 million (and then later 2 million) token input context length, and the ability to input video. I wrote about this at the time in The killer app of Gemini Pro 1.5 is video, which earned me a short appearance as a talking head in the Google I/O opening keynote in May.`
`How did the author contribute to the Google I/O opening keynote in May?`	`The earliest of those was Google’s Gemini 1.5 Pro, released in February. In addition to producing GPT-4 level outputs, it introduced several brand new capabilities to the field—most notably its 1 million (and then later 2 million) token input context length, and the ability to input video. I wrote about this at the time in The killer app of Gemini Pro 1.5 is video, which earned me a short appearance as a talking head in the Google I/O opening keynote in May.`
`How many organizations currently have models that rank higher than the original GPT-4 from March 2023?`	The GPT-4 barrier was comprehensively broken In my December 2023 review I wrote about how We don’t yet know how to build GPT-4—OpenAI’s best model was almost a year old at that point, yet no other AI lab had produced anything better. What did OpenAI know that the rest of us didn’t? I’m relieved that this has changed completely in the past twelve months. 18 organizations now have models on the Chatbot Arena Leaderboard that rank higher than the original GPT-4 from March 2023 (GPT-4-0314 on the board)—70 models in total.

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        768,
        512,
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 10
per_device_eval_batch_size: 10
num_train_epochs: 10
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 10
per_device_eval_batch_size: 10
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 1
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 5e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1
num_train_epochs: 10
max_steps: -1
lr_scheduler_type: linear
lr_scheduler_kwargs: {}
warmup_ratio: 0.0
warmup_steps: 0
log_level: passive
log_level_replica: warning
log_on_each_node: True
logging_nan_inf_filter: True
save_safetensors: True
save_on_each_node: False
save_only_model: False
restore_callback_states_from_checkpoint: False
no_cuda: False
use_cpu: False
use_mps_device: False
seed: 42
data_seed: None
jit_mode_eval: False
use_ipex: False
bf16: False
fp16: False
fp16_opt_level: O1
half_precision_backend: auto
bf16_full_eval: False
fp16_full_eval: False
tf32: None
local_rank: 0
ddp_backend: None
tpu_num_cores: None
tpu_metrics_debug: False
debug: []
dataloader_drop_last: False
dataloader_num_workers: 0
dataloader_prefetch_factor: None
past_index: -1
disable_tqdm: False
remove_unused_columns: True
label_names: None
load_best_model_at_end: False
ignore_data_skip: False
fsdp: []
fsdp_min_num_params: 0
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
tp_size: 0
fsdp_transformer_layer_cls_to_wrap: None
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
dataloader_pin_memory: True
dataloader_persistent_workers: False
skip_memory_metrics: True
use_legacy_prediction_loop: False
push_to_hub: False
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
eval_do_concat_batches: True
fp16_backend: auto
push_to_hub_model_id: None
push_to_hub_organization: None
mp_parameters:
auto_find_batch_size: False
full_determinism: False
torchdynamo: None
ray_scope: last
ddp_timeout: 1800
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
include_tokens_per_second: False
include_num_input_tokens_seen: False
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
eval_use_gather_object: False
average_tokens_across_devices: False
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	cosine_ndcg@10
1.0	16	0.9484
2.0	32	0.9455
3.0	48	0.9609
3.125	50	0.9609
4.0	64	0.9609
5.0	80	0.9638
6.0	96	0.9638
6.25	100	0.9638
7.0	112	0.9638
8.0	128	0.9638
9.0	144	0.9638
9.375	150	0.9638
10.0	160	0.9638

Framework Versions

Python: 3.11.12
Sentence Transformers: 4.1.0
Transformers: 4.51.3
PyTorch: 2.6.0+cu124
Accelerate: 1.6.0
Datasets: 3.5.1
Tokenizers: 0.21.1

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{henderson2017efficient,
    title={Efficient Natural Language Response Suggestion for Smart Reply},
    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},
    year={2017},
    eprint={1705.00652},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}