SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.04248 · 2024

Can Mamba Learn How to Learn? A Comparative Study on In-Context Learning Tasks

arXiv · PDF · Open in the Atlas

Code that ran

11 of 14 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, 1 violated it and 0 have no contract verdict; 9 failures of our instrument; 3 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/2402.04248.json · badge: /api/ran/2402.04248.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
krafton-ai/mambaformer-icl canonical (classifier label)
classifier basis: census
11 ran · 3 no recorded run
FunctionStatusWhere it lives
basic_plot Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/plot_utils.py
code served (permissive licence) · get_code("5b74687909baf467")
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 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 matplotlib.pyplot as plt
import seaborn as sns
import numpy as np

palette = sns.color_palette('colorblind')

def basic_plot(metrics, models=None, trivial=1.0, colors=None):
    fig, ax = plt.subplots(1, 1)
    if models is not None:
        metrics = {k: metrics[k] for k in models}
        if colors is not None:
            assert len(models) == len(colors)
            colors = {k: colors[i] for i, k in enumerate(models)}
    color = 0
    ax.axhline(trivial, ls='--', color='gray')
    for name, vs in metrics.items():
        ax.plot(vs['mean'], '-', label=name, color=colors[name] if colors else palette[color % 10], lw=2)
        low = vs['bootstrap_low']
        high = vs['bootstrap_high']
        ax.fill_between(range(len(low)), low, high, color=colors[name] if colors else None, alpha=0.2)
        color += 1
    ax.set_xlabel('in-context examples')
    ax.set_ylabel('squared error')
    ax.set_xlim(-1, len(low) + 0.1)
    ax.set_ylim(-0.1, 1.25)
    legend = ax.legend(loc='upper left', bbox_to_anchor=(1, 1))
    fig.set_size_inches(4, 3)
    for line in legend.get_lines():
        line.set_linewidth(3)
    return (fig, ax)

# Construct test data
metrics = {
    'model_a': {
        'mean': np.array([0.8, 0.6, 0.4, 0.3, 0.2]),
        'bootstrap_low': np.array([0.75, 0.55, 0.35, 0.25, 0.15]),
        'bootstrap_high': np.array([0.85, 0.65, 0.45, 0.35, 0.25])
    },
    'model_b': {
        'mean': np.array([0.9, 0.7, 0.5, 0.4, 0.3]),
        'bootstrap_low': np.array([0.85, 0.65, 0.45, 0.35, 0.25]),
        'bootstrap_high': np.array([0.95, 0.75, 0.55, 0.45, 0.35])
    }
}

models = ['model_a', 'model_b']
trivial = 1.0
colors = ['red', 'blue']

ARGS = {
    'metrics': metrics,
    'models': models,
    'trivial': trivial,
    'colors': colors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
matplotlib, numpy, seaborn — 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 float — 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
generate_dummy_vector Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/filtering_utils.py
code served (permissive licence) · get_code("b1c1ac424117b8de")
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 weight tensor with shape (2, 4)
# The function uses w.shape, w.device
# It generates random tensors matching w.shape and orthogonalizes them
w = torch.randn(2, 4)

ARGS = {
    "w": w
}
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
get_final_var Ran
ran, honoured the contract we drafted
krafton-ai/mambaformer-icl/src/curriculum.py
code served (permissive licence) · get_code("3c24ac252fef0868")
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 float, ndim 0, dtype family float, 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 math

ARGS = {
    "init_var": 0.1,
    "total_steps": 1000,
    "inc": 0.01,
    "n_steps": 100,
    "lim": 0.5
}
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 float, ndim 0, dtype family float, 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
mean_squared_error Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/tasks.py
code served (permissive licence) · get_code("fc94f27a0973e093")
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 0, dtype family float, 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 torch
import torch.nn as nn

torch.manual_seed(0)

# The function mean_squared_error takes two parameters:
# - ys_pred: predicted values (tensor)
# - ys: ground truth values (tensor)
# Both should be tensors that support .square() and .mean() operations

ys_pred = torch.randn(2, 4)
ys = torch.randn(2, 4)

ARGS = {
    "ys_pred": ys_pred,
    "ys": ys,
}
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 0, dtype family float, 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
orthogonalize Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/filtering_utils.py
code served (permissive licence) · get_code("a46d1c2a317b35c5")
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)

# Construct tiny tensors for orthogonalization
# vector: shape (batch_size, feature_dim)
# basis: shape (batch_size, feature_dim)
vector = torch.randn(2, 4)
basis = torch.randn(2, 4)

ARGS = {
    "vector": vector,
    "basis": basis,
}
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
sample_transformation Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/samplers.py
code served (permissive licence) · get_code("7f92ad21604fa243")
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 (4, 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
import math

torch.manual_seed(0)

# The function uses eigenvalues as a tensor that:
# - has len() called on it (so needs to be 1D)
# - is used in torch.diag()
# - is raised to power 2 and summed
# - is multiplied by a scalar

eigenvalues = torch.tensor([1.0, 2.0, 3.0, 4.0])

normalize = False

ARGS = {
    "eigenvalues": eigenvalues,
    "normalize": normalize,
}
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 (4, 4) — shape (4, 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
squared_error Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/tasks.py
code served (permissive licence) · get_code("348d63a78eb4cb94")
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
import torch.nn as nn

torch.manual_seed(0)

# Create small tensors for predicted and actual values
# batch_size=2, output_dim=4
ys_pred = torch.randn(2, 4)
ys = torch.randn(2, 4)

ARGS = {
    "ys_pred": ys_pred,
    "ys": ys,
}
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
squared_norm_error Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/tasks.py
code served (permissive licence) · get_code("b4326d9d300aa5c0")
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, 1) 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 small tensors for predictions and ground truth
ys_pred = torch.randn(2, 4)  # batch size 2, dimension 4
ys = torch.randn(2, 4)       # batch size 2, dimension 4

ARGS = {
    "ys_pred": ys_pred,
    "ys": ys,
    "dim": -1
}
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, 1) — shape (2, 1) 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
str2bool Ran
ran, violated the contract we drafted
krafton-ai/mambaformer-icl/src/util.py
code served (permissive licence) · get_code("8605dc8a088f3db8")
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) and returned bool, ndim 0, dtype family float, 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 argparse

def str2bool(v):
    if isinstance(v, bool):
        return v
    if v.lower() in ('yes', 'true', 't', 'y', '1'):
        return True
    elif v.lower() in ('no', 'false', 'f', 'n', '0'):
        return False
    else:
        raise argparse.ArgumentTypeError('Boolean value expected.')

# Test cases for the str2bool function
ARGS = {
    'v': 'true'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 bool, ndim 0, dtype family float, 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
train_step Ran
ran; our contract draft was wrong, not the code
krafton-ai/mambaformer-icl/src/train_parallel.py
code served (permissive licence) · get_code("47f61861b53edf4d")
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 torch.nn as nn

torch.manual_seed(0)

class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, xs, ys):
        # Return output with shape [batch, seq_len, 1] to trigger the ys_loss logic
        batch_size = xs.shape[0]
        return self.linear(xs).unsqueeze(-1)  # [2, 3, 1]

model = SimpleModel()
rank = 0
iter_num = 1
xs = torch.randn(2, 4)  # batch_size=2, feature_dim=4
ys = torch.randn(2, 3, 1)  # batch_size=2, seq_len=3, output_dim=1
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
loss_func = nn.MSELoss()
max_grad_norm = 1.0
filtering_inds = torch.tensor([False, False])  # No filtering

ARGS = {
    'rank': rank,
    'iter': iter_num,
    'model': model,
    'xs': xs,
    'ys': ys,
    'optimizer': optimizer,
    'loss_func': loss_func,
    'max_grad_norm': max_grad_norm,
    'filtering_inds': filtering_inds
}
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
zero_pad_ys Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
krafton-ai/mambaformer-icl/src/filtering_utils.py
code served (permissive licence) · get_code("75e826964a91f44b")
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 3, dtype family float, shape (2, 3, 1) 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)

# Case 1: tokenized=True
# When tokenized=True, xs.shape is (bsize, points) and dim=1
# ys should be (bsize, points) since it gets reshaped to (bsize, points, 1)
bsize = 2
points = 3

tokenized = True
xs_tokenized = torch.randn(bsize, points)
ys_tokenized = torch.randn(bsize, points)

# Case 2: tokenized=False
# When tokenized=False, xs.shape is (bsize, points, dim)
# ys should be (bsize, points) and will be reshaped/padded to (bsize, points, dim)
xs_not_tokenized = torch.randn(bsize, points, 4)
ys_not_tokenized = torch.randn(bsize, points)

# Test with tokenized=True
ARGS = {
    "tokenized": True,
    "xs": xs_tokenized,
    "ys": ys_tokenized
}
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 3, dtype family float, shape (2, 3, 1) — shape (2, 3, 1) 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
get_data_sampler Not yet run
unverified (harvested, no recorded run)
krafton-ai/mambaformer-icl/src/samplers.py
code served (permissive licence) · get_code("54f3966e15e1afe1")
How it ran

no invocation recorded.

randperm_2d_with_query Not yet run
unverified (harvested, no recorded run)
krafton-ai/mambaformer-icl/src/samplers.py
code served (permissive licence) · get_code("ddd35c92ba9c2871")
How it ran

no invocation recorded.

sample_seeds Not yet run
unverified (harvested, no recorded run)
krafton-ai/mambaformer-icl/src/train_parallel.py
code served (permissive licence) · get_code("4d49c710f74c99c4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 6 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 3 the same month, 6 later) and 8 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.

FunctionAlso present inEarliest-datedOthers named
str2bool 175 other papers
104 dated earlier, 8 the same month, 62 later, 1 same year, order not determinable · 2 only as an assembly brick
1406.6247 (2014-06)
Recurrent Models of Visual Attention
1603.09382 (2016-03)
Deep Networks with Stochastic Depth
1710.10196 (2017-10)
Progressive Growing of GANs for Improved Quality, Stability, and Variation
and 172 more not named here
get_final_var 8 other papers
2 dated earlier, 1 the same month, 5 later
2306.04891 (2023-06)
In-Context Learning through the Bayesian Prism
2310.17086 (2023-10)
Transformers Learn to Achieve Second-Order Convergence Rates for In-Context Linear Regression
2402.03170 (2024-02)
Is Mamba Capable of In-Context Learning?
and 5 more not named here
mean_squared_error 5 other papers
1 dated earlier, 4 later
2310.17086 (2023-10)
Transformers Learn to Achieve Second-Order Convergence Rates for In-Context Linear Regression
2404.18191 (2024-04)
Published at ICLR 2024 Workshop on Reliable and Responsible Foundation Models EXPLORING THE ROBUSTNESS OF IN-CONTEXT LEARNING WITH NOISY LABELS
2407.19346 (2024-07)
Polynomial Regression as a Task for Understanding In-context Learning Through Finetuning and Alignment
and 2 more not named here
squared_error 5 other papers
1 dated earlier, 4 later
2310.17086 (2023-10)
Transformers Learn to Achieve Second-Order Convergence Rates for In-Context Linear Regression
2404.18191 (2024-04)
Published at ICLR 2024 Workshop on Reliable and Responsible Foundation Models EXPLORING THE ROBUSTNESS OF IN-CONTEXT LEARNING WITH NOISY LABELS
2407.19346 (2024-07)
Polynomial Regression as a Task for Understanding In-context Learning Through Finetuning and Alignment
and 2 more not named here
sample_transformation 3 other papers
1 the same month, 2 later
2402.03170 (2024-02)
Is Mamba Capable of In-Context Learning?
2404.18191 (2024-04)
Published at ICLR 2024 Workshop on Reliable and Responsible Foundation Models EXPLORING THE ROBUSTNESS OF IN-CONTEXT LEARNING WITH NOISY LABELS
2509.07963 (2025-09)
Customizing the Inductive Biases of Softmax Attention using Structured Matrices
sample_seeds 2 other papers
1 dated earlier, 1 later
2310.17086 (2023-10)
Transformers Learn to Achieve Second-Order Convergence Rates for In-Context Linear Regression
2509.07963 (2025-09)
Customizing the Inductive Biases of Softmax Attention using Structured Matrices

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

State-space models (SSMs), such as Mamba (Gu & Dao, 2023), have been proposed as alternatives to Transformer networks in language modeling, by incorporating gating, convolutions, and input-dependent token selection to mitigate the quadratic cost of multi-head attention. Although SSMs exhibit competitive performance, their in-context learning (ICL) capabilities, a remarkable emergent property of modern language models that enables task execution without parameter optimization, remain underexplored compared to Transformers. In this study, we evaluate the ICL performance of SSMs, focusing on Mamba, against Transformer models across various tasks. Our results show that SSMs perform comparably to Transformers in standard regression ICL tasks, while outperforming them in tasks like sparse parity learning. However, SSMs fall short in tasks involving non-standard retrieval functionality. To address these limitations, we introduce a hybrid model, MambaFormer, that combines Mamba with attention blocks, surpassing individual models in tasks where they struggle independently. Our findings suggest that hybrid architectures offer promising avenues for enhancing ICL in language models.

For agents

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

get_harvested_code_for_paper("2402.04248")
get_code_for_paper("2402.04248")
have("2402.04248")

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

curl https://syntology.ai/api/ran/2402.04248.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.04248.svg)](https://syntology.ai/paper/2402.04248)

Connect an agent — have() is free.