Model Card for the Preference Optimization Interaction Baseline

A 124M model with the GPT-2 architecture trained for 20 "interaction rounds", using the training procedure outlined in the 2025 BabyLM Call for Papers.

Model Card for Preference Optimization Interaction Baseline
Table of Contents
Model Details
- Model Description
Uses
Training Details
Evaluation
- Testing Data & Metrics
  - Testing Data
  - Metrics
- Results
Technical Specifications
- Model Architecture and Objective
- Compute Infrastructure
Citation
Model Card Authors
Bibliography

Model Details

Model Description

This one of the two Interaction-track baselines 2025 BabyLM challenge.

Developed by: Mustafa Ömer Gül
Model type: Causal language model
Language(s) (NLP): eng
Resources for more information:
- GitHub Repo

Uses

This is a pre-trained language model. It can be used to evaluate tasks in a zero-shot manner and also can be fine-tuned for downstream tasks. It can be used for language generation but given its small size and low number of words trained on, do not expect LLM-level performance.

Training Details

Training Data

We used the BabyLM 100M (Strict) dataset to construct input contexts. It is composed of the following:

Source	Weight	Domain	Citation	Website	License
BNC	8%	Dialogue	BNC Consortium (2007)	link	link ¹
CHILDES	29%	Dialogue, Child-Directed	MacWhinney (2000)		link
Project Gutenberg	26%	Fiction, Nonfiction	Gerlach & Font-Clos (2020)	link	link
OpenSubtitles	20%	Dialogue, Scripted	Lison & Tiedermann (2016)	link	Open source
Simple English Wikipedia	15%	Nonfiction	--	link	link
Switchboard	1%	Dialogue	Godfrey et al. (1992), Stolcke et al., (2000)	link	link

¹ Our distribution of part of the BNC Texts is permitted under the fair dealings provision of copyright law (see term (2g) in the BNC license).

Hyperparameters

Hyperparameter	Value
Number of Rounds	20
Teacher model	Llama-3.1-8B-Instruct
Datapoint length	512
Context length	256
Student sampling temperature	1.0
Student top_p	0.8
Teacher sampling temperature	1.0
Teacher top_p	0.8
Pure language modeling epochs per round	8
Mixed language modeling + preference optimization epochs per round	2
Batch size	16
SimPO beta	2
SimPO gamma	0.5
Language modeling weight	0.2
Learning rate	0.00005
Number of steps	200000
Warmup steps	2000
Gradient clipping	1
Optimizer	AdamW
Optimizer Beta_1	0.9
Optimizer Beta_2	0.999
Optimizer Epsilon	10^-8

Training Procedure

The student model in this interaction baseline is trained for 20 "interaction rounds." Each round makes use of a distinct, randomly sampled 5M word subsample of the BabyLM corpus. Each interaction involves the student model generating a completion given an input context and the teacher model (Llama-3.1-8B-Instruct) producing a corrected version of the same completion. The model is trained with the standard next-token prediction loss on the concatenation of the context and teacher compleiton and with the SimPO loss on the teacher and student completions. More explicit details are available on the 2025 BabyLM call for papers.

Size and checkpoints

The model has 124M parameters. In total we train on around 1B words and provide multiple checkpoints from the training. Specifically we provode:

Checkpoints every 1M words for the first 10M words
Checkpoints every 10M words first 100M words
Checkpoints every 100M words until 1B words

Evaluation

This model is evaluated in three different fashions:

We do zero-shot evaluation on 7 tasks.
We do fine-tuning on a subset of the (Super)GLUE tasks (Wang et al., ICLR 2019; Wang et al., NeurIPS 2019) .

Testing Data & Metrics

Testing Data

For the BLiMP, BLiMP supplement, and EWoK tasks, we use a filtered version of the dataset to only include examples with words found in the BabyLM dataset. For the Finetuning task, we both filter and sample down to a maximum 10 000 train examples.

Validation Data

Zero-shot Tasks

BLiMP: The Benchmark of Linguistic Minimal Pairs evaluates the model's linguistic ability by seeing if it can recognize the grammatically correct sentence from a pair of minimally different sentences. It tests various grammatical phenomena.(Warstadt et al., TACL 2020)
BLiMP Supplement: A supplement to BLiMP introduced in the first edition of the BabyLM challenge. More focused on dialogue and questions. (Warstadt et al., CoNLL-BabyLM 2023)
EWoK: Works similarly to BLiMP but looks the model's internal world knowledge. Looking at both whter a model has physical and social knowledge. (Ivanova et al., 2024)
Eye Tracking and Self-paced Reading: Looks at whether the model can mimick the eye tracking and reading time of a human but using surprisal of a word as a proxy for time spent reading a word. (de Varda et al., BRM 2024)
Entity Tracking: Checks whether a model can keep track of the changes to the states of entities as text/dialogue unfolds. (Kim & Schuster, ACL 2023)
WUGs: Tests morphological generalization in LMs through an adjective nominalization task. (Hofmann et al., 2024)

Finetuning Tasks

BoolQ: A yes/no QA dataset with unprompted and unconstrained questions. (Clark et al., NAACL 2019)
MNLI: The Multi-Genre Natural Language Inference corpus tests the language understanding of a model by seeing wehther it can recognize textual entailment. (Williams et al., NAACL 2018)
MRPC: The Microsoft Research Paraphrase Corpus contains pairs of sentences that are either paraphrases/semntically equivalent to each other or unrelated.(Dolan & Brockett, IJCNLP 2005)
QQP²: Similarly to MRPC, the Quora Question Pairs corpus tests the models ability to determine whether a pair of questions are sematically similar to each other. These questions are sourced from Quora.
MultiRC: The Multi-Sentence Reading Comprehension corpus is a QA task that evaluates the model's ability to the correct answer from a list of answers given a question and context paragraph. In this version the data is changed to a binary classification judging whether the answer to a question, context pair is correct. (Khashabi et al., NAACL 2018)
RTE: Similar the Recognizing Text Entailement tests the model's ability to recognize text entailement. (Dagan et al., Springer 2006; Bar et al., 2006; Giampiccolo et al., 2007; Bentivogli et al., TAC 2009)
WSC: The Winograd Schema Challenge tests the models ability to do coreference resolution on sentences with a pronoun and a list of noun phrases found in the sentence. This version edits it to be a binary classification on examples consisting of a pronoun and noun phrase.(Levesque et al., PKRR 2012)

² https://www.quora.com/profile/Ricky-Riche-2/First-Quora-Dataset-Release-Question-Pairs

Metrics

The metrics used to evaluate the model are the following:

Zero-shot
- Accuracy on predicting the correct completion/sentence for BLiMP, BLiMP Supplement, EWoK, Entity Tracking, and WUGs
- Change in R^2 prediction from baseline for Eye Tracking (with no spillover) and Self-paced Reading (1-word spillover)
Finetuning
- 3 class Accuracy for MNLI
- Binary Accuracy for BoolQ, MultiRC, and WSC
- F1-score for MRPC and QQP

The metrics were chosen based on the advice of the papers the tasks come from.

Results

Zero-shot

Task	Metric	Causal Score
BLiMP	Acc	71.91
BLiMP Supplement	Acc	64.85
EWoK	Acc	52.44
Eye Tracking	change in R^2	0.5
Self-paced Reading	change in R^2	0.01
Entity Tracking	Acc	27.71
WUGs	Acc	38.5

Finetuning

Task	Metric	Uni-directional Score	Bi-directional Score
BoolQ	Acc
MNLI	Acc
MRPC	F1
QQP	F1
MultiRC	Acc
RTE	Acc
WSC	Acc

Technical Specifications

Hardware

1 H100 GPU was used to train this model.

Software

PyTorch

Training Time

The model took ~10-12 GPU hours to train, with a majority of this being spent on inference with the student and teacher models.

Citation

@misc{charpentier2025babylmturns3papers,
      title={BabyLM Turns 3: Call for papers for the 2025 BabyLM workshop}, 
      author={Lucas Charpentier and Leshem Choshen and Ryan Cotterell and Mustafa Omer Gul and Michael Hu and Jaap Jumelet and Tal Linzen and Jing Liu and Aaron Mueller and Candace Ross and Raj Sanjay Shah and Alex Warstadt and Ethan Wilcox and Adina Williams},
      year={2025},
      eprint={2502.10645},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2502.10645}, 
}