svsatheesh/BabySteps_SidBert-100M-mixed

BabySteps SidBert 100M (BabyLM 2026, strict track)

A GPT-BERT hybrid masked/causal language model trained on the official BabyLM 2026 strict corpus (100M words), submitted by team BabySteps to the 2026 BabyLM Challenge.

The central finding behind this model: under a fixed strong architecture at BabyLM scales, optimizer choice and learning-rate tuning outperformed every data-centric intervention we tried: curriculum-ordered corpora, textbook and story corpora, synthetic data generation, and teacher-supervised auxiliary losses. This model is a faithful GPT-BERT reconstruction trained with the Muon optimizer on the unmodified official corpus, trained on two gpus.

Model details

Architecture	GPT-BERT (hybrid masked + causal objective)
Parameters	~125M
Hidden size / layers / heads	768 / 12 / 12
Vocabulary	16,384 (BabyLM GPT-BERT baseline BPE)
Training sequence length	128
Objective	mixed (causal_per_masked = 1)
Optimizer	Muon (lr 0.005), auxiliary AdamW-style lr 0.007
Schedule	linear warmup + two-phase decay
Steps × tokens/step	12,330 × 131,072 (~1.6B tokens, ~12 epochs)
Precision	bf16
Seed	42
Hardware	2 × NVIDIA RTX 5090 (data-parallel)
Evaluation mode	MNTP

Training data

The official BabyLM 2026 strict corpus (100M words) — no custom, synthetic, or augmented data, and no teacher models.

Results (official BabyLM 2026 evaluation pipeline, MNTP)

Task	Score
BLiMP	78.69
BLiMP Supplement	72.68
EWoK	56.20
Entity Tracking	22.18
COMPS	57.66
Global PIQA	36.09
(Super)GLUE (finetuned)	72.42

SuperGLUE per task: BoolQ 71.8 · MNLI 66.2 · MRPC (F1) 90.7 · MultiRC 67.0 · QQP (F1) 77.4 · RTE 70.5 · WSC 63.5 Finetuning recipe: lr 2e-5, seed 42, batch size 4, 10 epochs.

Checkpoints

Intermediate checkpoints are published as branches following the BabyLM naming convention chck_{N}M, covering the required milestone schedule (1M–10M, 10M–100M, 100M–1000M words seen).

from transformers import AutoModelForMaskedLM, AutoTokenizer

model = AutoModelForMaskedLM.from_pretrained(
    "svsatheesh/BabySteps_SidBert-100M-mixed", trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("svsatheesh/BabySteps_SidBert-100M-mixed")

# a specific milestone checkpoint:
early = AutoModelForMaskedLM.from_pretrained(
    "svsatheesh/BabySteps_SidBert-100M-mixed", revision="chck_10M", trust_remote_code=True
)

The model can also be loaded with AutoModelForCausalLM, since GPT-BERT is trained under both objectives.

Reproducibility note

We trained several models with identical configuration and identical seed, differing only in GPU execution. Their scores span a full point of aggregate performance (seven-suite macro average 56.56–57.91, mean 57.31, sd 0.37), with per-suite ranges of 1.8 (BLiMP), 3.6 (BLiMP Supplement), 5.4 (entity tracking), 7.9 (Global PIQA) and 1.7 (SuperGLUE) points. Training is not bit-reproducible under bf16 with non-deterministic CUDA kernels, and small evaluation suites amplify the resulting differences. Point-scale differences between models at this scale should be interpreted with this variance in mind.

We additionally observe that two-GPU data-parallel runs of this configuration score systematically ~1 BLiMP point below single-GPU runs.

Limitations

Trained on 100M words of English only. Not instruction-tuned, not aligned, and not intended for production use — it is a research artifact for studying sample-efficient pretraining. Outputs may be ungrammatical, factually wrong, or reflect biases present in the BabyLM corpus.

Citation

@inproceedings{satheesh2026babysteps,
  title     = {BabySteps at BabyLM 2026: Optimizer Choice and Hyperparameter Tuning
               Beat Data Interventions for Sample-Efficient GPT-BERT Pretraining},
  author    = {Satheesh, Siddhi and Murillo, Jorge},
  booktitle = {Proceedings of the BabyLM Challenge at EMNLP 2026},
  year      = {2026}
}

Built on GPT-BERT (Charpentier & Samuel, 2024) and the Muon optimizer (Jordan et al., 2024).