This repository contains the Bi-Induct 1B Balanced checkpoint from Induction Signatures Are Not Enough: A Matched-Compute Study of Load-Bearing Structure in In-Context Learning.

This release corresponds to the 1B setting in the paper and is a research checkpoint intended for studying matched-compute pretraining, induction-style curricula, and in-context learning behavior. It is not instruction-tuned, alignment-tuned, or safety-tuned.

Variant

Bi-Induct balanced curriculum. Each synthetic injection chooses forward-copy or backward-copy with equal probability.

Model overview

Architecture: decoder-only Transformer
Positional encoding: RoPE (theta=10000)
Normalization: pre-norm residual blocks
MLP: SwiGLU
Attention: grouped-query / grouped key-value attention
Precision: bfloat16 training
Context length: 1024
Embeddings: untied input/output embeddings

Model specification

Field	Value
Parameters (paper label)	1B
Layers	30
Hidden size	1,536
Intermediate / MLP size	6,144
Head dimension	64
Attention heads	24
KV heads	6

Training data

All checkpoints in this family were pretrained on the deduplicated THE PILE in streaming / shuffled mode. A stable MD5-based hash was used to create a fixed held-out evaluation slice, with 0.2% of the corpus reserved for evaluation (roughly 0.4B tokens). Tokenization was truncated to 1024 tokens per sequence.

For the Bi-Induct variants, synthetic snippets were interleaved on top of the natural stream:

Induction: [S || SEP || S]
Anti-Induction: [S || SEP || reverse(S)]
Balanced: each injection randomly chooses induction or anti-induction

The main cross-scale experiments used span length L = 20 and initial mix ratio m0 = 50%, linearly annealed to zero over the full training budget.

Training recipe

Optimizer: AdamW (beta1=0.9, beta2=0.999, weight decay 0.1)
Learning rate: peak 1e-3
Schedule: 3% linear warmup, then cosine decay
Update size: 2^16 tokens per update
Token budget: approximately 20N tokens following the Chinchilla-style rule of thumb
Comparison protocol: iso-FLOPs across curricula at each scale

Evaluation summary for the 1B family

The table below summarizes the main results at this scale. Standard LM benchmarks are evaluated 3-shot and Todd et al. function-style probes are evaluated 10-shot with HITS@1.

Variant	Standard LM ICL composite ↑	Todd-style ICL composite ↑	Held-out PPL ↓
Baseline	24.2 ± 0.5	20.0 ± 1.3	14.1
Induction	23.9 ± 0.5	15.2 ± 1.1	14.9
Anti-Induction	23.6 ± 0.4	14.7 ± 1.2	14.9
Balanced	24.3 ± 0.3	14.9 ± 1.1	14.9

This checkpoint: Balanced.

Benchmarks included

Standard LM benchmarks

MMLU
Winogrande
CommonSenseQA
PIQA
HellaSwag
TriviaQA-Wiki
BBH (CoT)
OpenBookQA
ARC-Challenge
GPQA
GSM-8K
MathQA
BoolQ
LAMBADA

Todd et al. function-style probes

alphabetically first 3
alphabetically first 5
alphabetically last 3
alphabetically last 5
capitalize
capitalize first letter
capitalize last letter
choose first of 3
choose first of 5
choose last of 3
choose last of 5
choose middle of 3
choose middle of 5
lowercase first letter
lowercase last letter
next capital letter
next item
prev item
word length

Example usage

python

from transformers import AutoTokenizer, AutoModelForCausalLM

repo_id = "MohammedSabry/biinduct-1b-balanced"

tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id)

prompt = "The capital of France is"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=20)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Limitations

These are research checkpoints, not production chat models.
They were designed to study the relationship between induction-style telemetry and load-bearing ICL behavior under matched compute.
The synthetic interventions are intentionally lightweight and token-level; results should not be interpreted as ruling out richer data-rewrite strategies.
Because Bi-Induct replaces a fraction of natural data under iso-FLOPs, some trade-offs may reflect natural-text displacement in addition to mechanistic redundancy.

Citation

If you use this model, please cite:

bibtex

@misc{sabry2026inductionsignaturesenoughmatchedcompute,
      title={Induction Signatures Are Not Enough: A Matched-Compute Study of Load-Bearing Structure in In-Context Learning}, 
      author={Mohammed Sabry and Anya Belz},
      year={2026},
      eprint={2509.22947},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2509.22947}, 
}