Netta1994/setfit_baai_cybereason_gpt-4o_improved-cot-instructions_chat_few_shot_generated_remov

SetFit with BAAI/bge-base-en-v1.5

This is a SetFit model that can be used for Text Classification. This SetFit model uses BAAI/bge-base-en-v1.5 as the Sentence Transformer embedding model. A LogisticRegression instance is used for classification.

The model has been trained using an efficient few-shot learning technique that involves:

1. Fine-tuning a Sentence Transformer with contrastive learning.
2. Training a classification head with features from the fine-tuned Sentence Transformer.

Model Details

Model Description

• Model Type: SetFit
• Sentence Transformer body: BAAI/bge-base-en-v1.5
• Classification head: a LogisticRegression instance
• Maximum Sequence Length: 512 tokens
• Number of Classes: 2 classes

Model Sources

• Repository: SetFit on GitHub
• Paper: Efficient Few-Shot Learning Without Prompts
• Blogpost: SetFit: Efficient Few-Shot Learning Without Prompts

Model Labels

Label	Examples
1	'Reasoning:\nThe given answer is directly aligned with the context provided by the document. It accurately identifies Ennita Manyumwa as a 26-year-old African woman who is HIV-free. It further elaborates on her significance in the fight against AIDS by highlighting her resistance to older men who offer gifts for sex, a behavior that helps prevent the spread of AIDS. This information is consistent with the document and directly answers the question without any unnecessary details.\n\nFinal Evaluation:' 'Reasoning:\nThe answer directly addresses the question by listing the benefits the author has experienced from their regular yoga practice. These benefits include unapologetic "me" time, improved health, self-growth, increased patience, the ability to be still, acceptance of daily changes, the realization that happiness is their responsibility, a deeper appreciation for their body, the understanding that yoga exists off the mat, and the importance of being open. Each of these points is explicitly mentioned in the provided document, making the answer well-supported and contextually accurate. The answer is concise and relevant, sticking closely to the specifics asked for in the question.\n\nFinal Evaluation:' 'Reasoning:\nThe answer accurately identifies that the work on germ-free-life conducted at Notre Dame University resulted in the establishment of the Lobund Institute. This directly aligns with the information provided in the document, which details the evolution of the research from its beginnings in 1928 to the establishment of the Lobund Institute. The response is relevant, well-grounded in the context of the document, and concise.\n\nEvaluation:'
0	'Reasoning:\nThe answer provided accurately addresses the question by explaining how to enable approval for appointment bookings, which subsequently changes the booking process for clients from immediate booking to a "request to book" process. This may be a solution to the issue if clients are currently experiencing difficulties due to the lack of this feature. The steps given are clear, concise, and directly supported by the provided document, aligning well with the instructions mentioned for enabling approval.\n\nHowever, it is important to note that the answer does not directly state why clients might be unable to book appointments online, nor does it explore other potential reasons beyond the approval setting. Directly stating that clients cannot book appointments online due to lack of enabling approval, and covering any other potential issues mentioned in the document, would make it even more thorough.\n\nEvaluation:' 'Reasoning:\nThe answer does cover the fundamental steps to write a nonfiction book, such as selecting a topic, conducting research, creating an outline, and starting the writing process. However, it includes an incorrect aside, stating that "The Old Man and the Sea" by Ernest Hemingway is nonfiction and based on true events, which detracts from the otherwise accurate guidance. Additionally, the answer could be more detailed and aligned with the extensive steps provided in the document, such as discussing the importance of understanding the genre, reading and analyzing examples, brainstorming, setting up interviews, organizing research, creating a writing schedule, and focusing on writing techniques.\n\nFinal Evaluation: \nEvaluation:' 'Reasoning:\nThe provided answer directly contradicts the guidelines given in the document on studying English literature. The answer suggests not taking notes, ignoring significant passages, and avoiding making character profiles, which are all contrary to the recommendations in the document. The document emphasizes the importance of thorough reading, taking detailed notes, creating character profiles, and paying attention to important passages and concepts, which are crucial for comprehensive understanding and analysis of English literature.\n\nFinal Evaluation: \nEvaluation:'

Evaluation

Metrics

Label	Accuracy
all	0.6761

Uses

Direct Use for Inference

First install the SetFit library:

pip install setfit

Then you can load this model and run inference.

from setfit import SetFitModel

# Download from the 🤗 Hub
model = SetFitModel.from_pretrained("Netta1994/setfit_baai_cybereason_gpt-4o_improved-cot-instructions_chat_few_shot_generated_remov")
# Run inference
preds = model("Reasoning:
The provided document clearly outlines the purpose of the <ORGANIZATION> XDR On-Site Collector Agent: it is installed to collect logs from platforms and securely forward them to <ORGANIZATION> XDR. The answer given aligns accurately with the document's description, addressing the specific question without deviating into unrelated topics. The response isalso concise and to the point.

Evaluation:")

Training Details

Training Set Metrics

Training set	Min	Median	Max
Word count	33	96.1280	289

Label	Training Sample Count
0	312
1	321

Training Hyperparameters

• batch_size: (16, 16)
• num_epochs: (2, 2)
• max_steps: -1
• sampling_strategy: oversampling
• num_iterations: 20
• body_learning_rate: (2e-05, 2e-05)
• head_learning_rate: 2e-05
• loss: CosineSimilarityLoss
• distance_metric: cosine_distance
• margin: 0.25
• end_to_end: False
• use_amp: False
• warmup_proportion: 0.1
• l2_weight: 0.01
• seed: 42
• eval_max_steps: -1
• load_best_model_at_end: False

Training Results

Epoch	Step	Training Loss	Validation Loss
0.0006	1	0.2154	-
0.0316	50	0.2582	-
0.0632	100	0.2517	-
0.0948	150	0.2562	-
0.1263	200	0.2532	-
0.1579	250	0.2412	-
0.1895	300	0.184	-
0.2211	350	0.1608	-
0.2527	400	0.1487	-
0.2843	450	0.117	-
0.3159	500	0.0685	-
0.3474	550	0.0327	-
0.3790	600	0.0257	-
0.4106	650	0.0139	-
0.4422	700	0.012	-
0.4738	750	0.0047	-
0.5054	800	0.0046	-
0.5370	850	0.0042	-
0.5685	900	0.0058	-
0.6001	950	0.0029	-
0.6317	1000	0.0055	-
0.6633	1050	0.0033	-
0.6949	1100	0.0026	-
0.7265	1150	0.0026	-
0.7581	1200	0.0033	-
0.7896	1250	0.0049	-
0.8212	1300	0.0043	-
0.8528	1350	0.0019	-
0.8844	1400	0.0015	-
0.9160	1450	0.0014	-
0.9476	1500	0.0017	-
0.9792	1550	0.0013	-
1.0107	1600	0.0019	-
1.0423	1650	0.0012	-
1.0739	1700	0.0011	-
1.1055	1750	0.0013	-
1.1371	1800	0.0012	-
1.1687	1850	0.0013	-
1.2003	1900	0.0013	-
1.2318	1950	0.0012	-
1.2634	2000	0.0011	-
1.2950	2050	0.0012	-
1.3266	2100	0.0011	-
1.3582	2150	0.0011	-
1.3898	2200	0.0012	-
1.4214	2250	0.0014	-
1.4529	2300	0.0011	-
1.4845	2350	0.001	-
1.5161	2400	0.0011	-
1.5477	2450	0.001	-
1.5793	2500	0.001	-
1.6109	2550	0.0012	-
1.6425	2600	0.0011	-
1.6740	2650	0.0011	-
1.7056	2700	0.001	-
1.7372	2750	0.001	-
1.7688	2800	0.001	-
1.8004	2850	0.001	-
1.8320	2900	0.001	-
1.8636	2950	0.001	-
1.8951	3000	0.001	-
1.9267	3050	0.0009	-
1.9583	3100	0.0011	-
1.9899	3150	0.001	-

Framework Versions

• Python: 3.10.14
• SetFit: 1.1.0
• Sentence Transformers: 3.1.1
• Transformers: 4.44.0
• PyTorch: 2.4.0+cu121
• Datasets: 3.0.0
• Tokenizers: 0.19.1

Citation

BibTeX

@article{https://doi.org/10.48550/arxiv.2209.11055,
    doi = {10.48550/ARXIV.2209.11055},
    url = {https://arxiv.org/abs/2209.11055},
    author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren},
    keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences},
    title = {Efficient Few-Shot Learning Without Prompts},
    publisher = {arXiv},
    year = {2022},
    copyright = {Creative Commons Attribution 4.0 International}
}