SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.14955 · 2024

ICLEval: Evaluating In-Context Learning Ability of Large Language Models

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 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; 7 failures of our instrument; 2 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/2406.14955.json · badge: /api/ran/2406.14955.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
yiye3/icleval canonical (classifier label)
classifier basis: census
8 ran · 1 no recorded run
rucbm/icleval canonical (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
check_order Ran
ran; our contract draft was wrong, not the code
yiye3/icleval/code/generate_data/learning_ability/classifier_order.py
code served (permissive licence) · get_code("925d81f4d1098c6e")
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 dict. 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 random

random.seed(0)

# Analyze the function:
# - candidates: used with random.sample(candidates, 5), so needs to be a sequence with at least 5 elements
# - uid: used as a dict value, can be any identifier
# - task_type: used as a dict value, can be a string
# - n_shot: integer parameter, default 8
# - task: string parameter, default 'check_order'

# The function doesn't actually use candidates in a way that requires strings,
# but the code calls ', '.join(elements) so candidates should be strings
candidates = ['apple', 'banana', 'cherry', 'date', 'elderberry', 'fig', 'grape', 'honeydew']
uid = 'user_123'
task_type = 'sequence_reversal'
n_shot = 2  # small value for testing
task = 'check_order'

ARGS = {
    'candidates': candidates,
    'uid': uid,
    'task_type': task_type,
    'n_shot': n_shot,
    'task': task
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
check_repeated_content Ran
ran; our contract draft was wrong, not the code
yiye3/icleval/code/generate_data/learning_ability/classifier_duplication.py
code served (permissive licence) · get_code("028f2fcce9aa8609")
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 dict. 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 random

# Set seed for reproducibility
random.seed(42)

# Define arguments for check_repeated_content
ARGS = {
    'candidates': ['apple', 'banana', 'cherry', 'date', 'elderberry', 'fig', 'grape', 'honeydew'],
    'uid': 'user_123',
    'task_type': 'sentence',
    'n_shot': 8,
    'task': 'check_repeated_content'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
dict_search_number Ran
ran; our contract draft was wrong, not the code
yiye3/icleval/code/generate_data/copying_ability/dict_search_number.py
code served (permissive licence) · get_code("a4303b56a1de0b68")
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 dict. 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 random

random.seed(0)

ARGS = {
    'uid': 'test_uid_001',
    'task_type': 'number-all_similar'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
dict_search_string Ran
ran; our contract draft was wrong, not the code
yiye3/icleval/code/generate_data/copying_ability/dict_search_string.py
code served (permissive licence) · get_code("1fa76eb617bf97a2")
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 dict. 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 random
import hashlib

random.seed(42)

uid = "user_123"
sentences = [
    "The quick brown fox jumps over the lazy dog",
    "Machine learning is a subset of artificial intelligence",
    "Python is a versatile programming language",
    "Data science combines statistics and programming",
    "Neural networks are inspired by biological neurons",
    "Deep learning has revolutionized computer vision",
    "Natural language processing enables text understanding",
    "Transformers have become state of the art in NLP",
    "Large language models show emergent capabilities",
    "Attention mechanisms are crucial for sequence models",
    "Gradient descent optimizes neural network weights",
    "Backpropagation computes gradients efficiently",
    "Regularization prevents overfitting in models",
    "Cross validation estimates generalization error",
    "Feature engineering improves model performance",
    "Hyperparameter tuning optimizes model behavior",
    "Batch normalization stabilizes training",
    "Dropout adds noise for better generalization",
    "Embeddings represent words as dense vectors",
    "Word2vec learns meaningful word representations",
    "Convolutional networks process spatial data",
    "Recurrent networks handle sequential data",
    "LSTM cells help with long term dependencies",
    "Attention weights focus on relevant information"
]

ARGS = {
    "uid": uid,
    "sentences": sentences
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
find_mha_dims Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
yiye3/icleval/code/model_evaluation/modeling_phi/modeling_mixformer_sequential.py
code served (permissive licence) · get_code("b93732e7a9dd5f8e")
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, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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
from transformers import PretrainedConfig

class MockConfig(PretrainedConfig):
    """Minimal config class with required attributes for find_mha_dims."""
    model_type = "mock"
    
    def __init__(self, n_embd: int = 64, n_head: int = 8, **kwargs):
        super().__init__(**kwargs)
        self.n_embd = n_embd
        self.n_head = n_head

# Create a valid config with n_embd divisible by n_head
config = MockConfig(n_embd=64, n_head=8)

ARGS = {
    "config": config,
    "n_head": None,
    "head_dim": None,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
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
get_similar_number Ran
ran, honoured the contract we drafted
yiye3/icleval/code/generate_data/copying_ability/dict_search_number.py
code served (permissive licence) · get_code("22644c9d1faa3ab4")
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 int, ndim 0, dtype family int, shape () 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 random

def get_similar_number(number):
    number = [x for x in str(number)]
    pos_id = random.sample(range(len(number)), 1)[0]
    number[pos_id] = str(random.sample(range(10), 1)[0])
    return int(''.join(number))

ARGS = {
    "number": 12345
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () 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
load_one_sample Ran
ran; our contract draft was wrong, not the code
yiye3/icleval/code/generate_data/copying_ability/dict_search_number.py
code served (permissive licence) · get_code("77d951c66faed420")
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 list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 random

# Set seed for reproducibility
random.seed(42)

# ARGS dict with all parameters for load_one_sample
ARGS = {
    'sim_num': 5,
    'mode': 'long'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
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
natural_language_string Ran
ran; our contract draft was wrong, not the code
yiye3/icleval/code/generate_data/copying_ability/natural_language_string.py
code served (permissive licence) · get_code("211c6dd070d2fd05")
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 dict. 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 hashlib

# Construct a line dict with the required keys: 'title', 'content', 'count'
line = {
    'title': 'Hello',
    'content': 'This is Hello world and Hello again',
    'count': 5
}

# Construct uid parameter
uid = 'uid_001'

ARGS = {
    'uid': uid,
    'line': line
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not 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
check_format Not yet run
unverified (harvested, no recorded run)
yiye3/icleval/code/generate_data/learning_ability/classifier_format.py
code served (permissive licence) · get_code("6f2cc86ddb8c2b5b")
How it ran

no invocation recorded.

greedy_until Not yet run
unverified (harvested, no recorded run)
rucbm/icleval/code/model_evaluation/run_icl_eval.py
code served (permissive licence) · get_code("6dd3de7ed4a93b10")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 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 10 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

In-Context Learning (ICL) is a critical capability of Large Language Models (LLMs) as it empowers them to comprehend and reason across interconnected inputs. Evaluating the ICL ability of LLMs can enhance their utilization and deepen our understanding of how this ability is acquired at the training stage. However, existing evaluation frameworks primarily focus on language abilities and knowledge, often overlooking the assessment of ICL ability. In this work, we introduce the ICLEval benchmark to evaluate the ICL abilities of LLMs, which encompasses two key sub-abilities: exact copying and rule learning. Through the ICLEval benchmark, we demonstrate that ICL ability is universally present in different LLMs, and model size is not the sole determinant of ICL efficacy. Surprisingly, we observe that ICL abilities, particularly copying, develop early in the pretraining process and stabilize afterward. Our source codes and benchmark are released at https://github.com/yiye3/ICLEval.

For agents

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

get_harvested_code_for_paper("2406.14955")
get_code_for_paper("2406.14955")
have("2406.14955")

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

curl https://syntology.ai/api/ran/2406.14955.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.14955.svg)](https://syntology.ai/paper/2406.14955)

Connect an agent — have() is free.