StanfordSCALE/assertion_sentence_evaluates_a_student_response

Assertion: sentence evaluates a student response

This classifier was trained for EduBehaviors: Assertion-based schemas for auditable dialogue coding and is usable through the Python package EduBehaviors-kit. This classifier was trained on an LLM-annotated subset of teacher utterances from the TalkMoves Dataset. See the Datasets section below for more information.

---

Training Details

Datasets

Dataset	Split	Size
StanfordSCALE/assertions_llm_annotated_talkmoves	train	3,430 (53.3%)
StanfordSCALE/assertions_llm_annotated_talkmoves	dev	860 (13.4%)
StanfordSCALE/assertions_llm_annotated_talkmoves	test	2,144 (33.3%)

This model's columns are assertion_sentence_evaluates_a_student_response and split_sentence_evaluates_a_student_response.

Base rate (share of rows labeled as True): 6.5% overall — 6.2% train, 6.4% dev, 7.0% test.

Labels and annotation

Labels were generated with LLM annotators. Krippendorff's alpha for this assertion is 0.433.

Hyperparameters

Parameter	Value
Base model (body)	sentence-transformers/paraphrase-mpnet-base-v2
Head	LogisticRegression
Body learning rate	2e-05
Head learning rate	0.01
Batch size	16 (contrastive phase) / 32 (head)
Epochs	10
Max steps	5000 (contrastive phase)
Eval max steps	100
Seed	20260904
Mixed precision	enabled on GPU

---

Evaluation

Results

Split	n	Base rate	Precision	Recall	F1 (positive class)	ROC-AUC	Average precision
dev	860	6.4%	0.588	0.364	0.449	0.855	0.501
test	2,144	7.0%	0.472	0.333	0.391	0.815	0.374

Limitations

• Labels come from LLM annotators, not human coders. Agreement between annotators with Krippendorff's Alpha is 0.433.
• Trained on teacher utterances only. Behaviour on student speech is untested.
• Test F1 is 0.391. This model does not work well enough to be used on its own.

---

How to Use

Message Structure

The model was trained on text built as:

{utterance}

The utterance is passed through as-is.

Running instructions

pip install setfit

from setfit import SetFitModel

model = SetFitModel.from_pretrained("StanfordSCALE/assertion_sentence_evaluates_a_student_response")

text = 'Cool'
model.predict([text])        # -> array([1]) when the assertion holds
model.predict_proba([text])  # -> [[P(no), P(yes)]]

Citation

@misc{assertion_sentence_evaluates_a_student_response,
  author = {Stanford SCALE Initiative},
  title  = {Assertion classifier: sentence evaluates a student response},
  year   = {2026},
  url    = {https://huggingface.co/StanfordSCALE/assertion_sentence_evaluates_a_student_response}
}