SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.02550 · NeurIPS · 2024

Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks

Tianyu He, Andrey Gromov, Darshil Doshi, Aritra Das

arXiv · PDF · Open in the Atlas

Code that ran

3 of 3 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 0 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2406.02550.json · badge: /api/ran/2406.02550.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
ablghtianyi/ICL_Modular_Arithmetic canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
pca Ran
ran; our contract draft was wrong, not the code
ablghtianyi/ICL_Modular_Arithmetic/interpretability/attn_map.py
code served (permissive licence) · get_code("210f105fe6762eeb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small embedding tensor
# Shape: (n_embd, vocab_size) = (4, 3)
# The function expects embedding.shape to be (n_embd, vocab_size)
embedding = torch.randn(4, 3)

# n_top_direction parameter
n_top_direction = 2

ARGS = {
    "embedding": embedding,
    "n_top_direction": n_top_direction,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
scaled_dot_product_attention Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ablghtianyi/ICL_Modular_Arithmetic/interpretability/attn_map.py
code served (permissive licence) · get_code("4f3bdb6c844ac5cc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import math

torch.manual_seed(0)

# Construct small tensors for query, key, value
# query: (batch=2, seq_len=4, dim=8)
# key: (batch=2, seq_len=3, dim=8)
# value: (batch=2, seq_len=3, dim=6)
query = torch.randn(2, 4, 8)
key = torch.randn(2, 3, 8)
value = torch.randn(2, 3, 6)

# attn_mask: optional, can be None or a boolean/float mask
# If boolean: (L, S) where L=4, S=3
attn_mask = None

# dropout_p: float between 0 and 1
dropout_p = 0.1

# is_causal: boolean flag
is_causal = True

# scale: optional float scale factor
scale = None

ARGS = {
    "query": query,
    "key": key,
    "value": value,
    "attn_mask": attn_mask,
    "dropout_p": dropout_p,
    "is_causal": is_causal,
    "scale": scale,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
svd_flip Ran
ran, honoured the contract we drafted
ablghtianyi/ICL_Modular_Arithmetic/interpretability/attn_map.py
code served (permissive licence) · get_code("ae0b4e61e0a7103b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create small tensors for u and v
# u should be shape (m, n) where we'll take argmax along axis 0
# v should be shape (n, k) to be compatible with the signs multiplication
u = torch.randn(4, 3)  # 4 rows, 3 columns
v = torch.randn(3, 2)  # 3 rows, 2 columns (distinct from u's dimensions)

ARGS = {
    "u": u,
    "v": v,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 3 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions z = a x + b y mod p labeled by the vector (a, b) ∈ Z 2 p . We use some of these tasks for pre-training and the rest for out-ofdistribution testing. We empirically show that a GPT-style transformer exhibits a transition from in-distribution to out-of-distribution generalization as the number of pre-training tasks increases. We find that the smallest model capable of outof-distribution generalization requires two transformer blocks, while for deeper models, the out-of-distribution generalization phase is transient, necessitating early stopping. Finally, we perform an interpretability study of the pre-trained models, revealing highly structured representations in both attention heads and MLPs; and discuss the learned algorithms. Notably, we find an algorithmic shift in deeper models, as we go from few to many in-context examples.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2406.02550")
get_code_for_paper("2406.02550")
have("2406.02550")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2406.02550.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2406.02550.svg)](https://syntology.ai/paper/2406.02550)

Connect an agent — have() is free.