SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2601.16631 · 2026

PanopMamba: Vision State Space Modeling for Nuclei Panoptic Segmentation

Ming Kang, C.-W Phan, Chee-Ming Ting, Fung Fung, F Kang, R Ting, -W Phan, C.-M Ting

arXiv · PDF · Open in the Atlas

Code that ran

13 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (12 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 13 have no contract verdict; 0 failures of our instrument; 4 no recorded run), measured 2026-11-22; 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/2601.16631.json · badge: /api/ran/2601.16631.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
mkang315/PanopMamba community
Papers with Code archive
13 ran · 4 no recorded run
FunctionStatusWhere it lives
ConvFFN Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("d3c570751b908509")
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 an instance of its own class, ConvFFN: the run constructed the object and did not call it, so no result was computed. 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)

ARGS = {
    "channels": 16,
    "expansion": 2,
    "drop": 0.2,
}
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 ConvFFN — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a ConvFFN 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
CrossMerge Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("bf30be820b2faced")
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 an instance of its own class, CrossMerge: the run constructed the object and did not call it, so no result was computed. 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)

# CrossMerge.forward takes a tensor ys with shape (B, K, D, H, W)
# where K should be 4 (since it uses indices 0:2 and 2:4)
# Let's use distinct sizes: B=2, K=4, D=3, H=5, W=6

B, K, D, H, W = 2, 4, 3, 5, 6
ys = torch.randn(B, K, D, H, W)

ARGS = {
    '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 CrossMerge — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CrossMerge 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
CrossMerge2 Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("4d41a84773237a25")
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 an instance of its own class, CrossMerge2: the run constructed the object and did not call it, so no result was computed. 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)

# CrossMerge2.forward expects a tensor ys with shape (B, K, D, H, W)
# where:
# - B is batch size
# - K is 2 (it accesses ys[:, 0] and ys[:, 1])
# - D is feature dimension
# - H, W are spatial dimensions
# Let's use: B=2, K=2, D=4, H=3, W=5

ys = torch.randn(2, 2, 4, 3, 5)

ARGS = {
    "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 CrossMerge2 — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CrossMerge2 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
CrossMerge3 Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("4f34d80bc5c2ae69")
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 an instance of its own class, CrossMerge3: the run constructed the object and did not call it, so no result was computed. 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)

# CrossMerge3.forward expects a tensor ys of shape (B, K, D, H, W)
# where B=batch, K=2 (appears to be used with indices 0, 1, 2), D=depth, H=height, W=width
# CrossMerge3.backward expects the gradient to have shape (B, C, L) where L = H*W

B = 2
K = 3
D = 4
H = 3
W = 5

# Create the input tensor with shape (B, K, D, H, W)
ys = torch.randn(B, K, D, H, W)

ARGS = {
    "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 CrossMerge3 — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CrossMerge3 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
CrossScan Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("930e2c172dd437df")
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 an instance of its own class, CrossScan: the run constructed the object and did not call it, so no result was computed. 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)

# CrossScan.forward expects:
# - x: torch.Tensor with shape (B, C, H, W)
# The function flattens and transposes this tensor in various ways

# Create a small test tensor with distinct dimensions for debugging
# B=2 (batch), C=4 (channels), H=3 (height), W=5 (width)
x = torch.randn(2, 4, 3, 5)

ARGS = {
    "x": x,
}
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 CrossScan — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CrossScan 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
CrossScan2 Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("2e083da1d636fbfd")
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 an instance of its own class, CrossScan2: the run constructed the object and did not call it, so no result was computed. 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)

# CrossScan2.forward takes a single tensor argument x with shape (B, C, H, W)
# The function reshapes and transposes this tensor
# We need a small 4D tensor with distinct dimension sizes

x = torch.randn(2, 4, 3, 5)  # B=2, C=4, H=3, W=5

ARGS = {
    "x": x
}
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 CrossScan2 — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CrossScan2 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
CrossScan3 Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("05f3d40ee609910b")
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 an instance of its own class, CrossScan3: the run constructed the object and did not call it, so no result was computed. 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)

# Construct a small input tensor with shape (B, C, H, W)
# B=2 (batch), C=4 (channels), H=3 (height), W=5 (width)
x = torch.randn(2, 4, 3, 5)

ARGS = {
    "x": x,
}
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 CrossScan3 — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CrossScan3 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
Mlp Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("65c1e05eec0dfce5")
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 an instance of its own class, Mlp: the run constructed the object and did not call it, so no result was computed. 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
from functools import partial

torch.manual_seed(0)

ARGS = {
    "in_features": 4,
    "hidden_features": 8,
    "out_features": 3,
    "act_layer": nn.GELU,
    "drop": 0.1,
    "channels_first": False,
}
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 Mlp — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Mlp 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
PatchMerging2D Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("0c300d0d98039a0f")
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 an instance of its own class, PatchMerging2D: the run constructed the object and did not call it, so no result was computed. 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)

ARGS = {
    "dim": 4,
    "out_dim": -1,
    "norm_layer": nn.LayerNorm,
}
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 PatchMerging2D — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a PatchMerging2D 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
SEModule Ran
ran on a synthesized input
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("15c4c79446f1db10")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, 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 torch
import torch.nn as nn

INIT_ARGS = {
    'channels': 4,
    'reduction': 2,
    'act_layer': nn.ReLU,
    'gate_layer': 'sigmoid'
}

ARGS = {
    'x': torch.randn(2, 4, 3, 5)
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 4, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
SS2D Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("4f2d0071a8dc9ec4")
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 an instance of its own class, SS2D: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import numpy as np

torch.manual_seed(0)

# Build minimal tensors for SS2D forward pass
B, C, H, W = 2, 4, 3, 3  # batch, channels, height, width
L = H * W  # sequence length

# u: input tensor (B, C, H, W)
u = torch.randn(B, C, H, W, dtype=torch.float32)

# delta: step size tensor (B, C, H, W)
delta = torch.randn(B, C, H, W, dtype=torch.float32)

# A: state transition matrix (C, L) or (1, L)
A = torch.randn(C, L, dtype=torch.float32)

# B: input-to-state matrix (B, L) or (B, 1, L)
B_tensor = torch.randn(B, 1, L, dtype=torch.float32)

# C: output matrix (B, L) or (B, 1, L)
C_tensor = torch.randn(B, 1, L, dtype=torch.float32)

# D: feedthrough (C,) or scalar
D = torch.randn(C, dtype=torch.float32)

# delta_bias: bias for delta (C,)
delta_bias = torch.randn(C, dtype=torch.float32)

# delta_softplus: boolean
delta_softplus = True

# nrows: number of rows for scan
nrows = 1

ARGS = {
    'u': u,
    'delta': delta,
    'A': A,
    'B': B_tensor,
    'C': C_tensor,
    'D': D,
    'delta_bias': delta_bias,
    'delta_softplus': delta_softplus,
    'nrows': nrows,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SS2D — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SS2D 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
SelectiveScan Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("b76893328a17f33e")
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 an instance of its own class, SelectiveScan: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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)

# Mock the CUDA selective scan modules since they won't be available
class MockSelectiveScanCuda:
    @staticmethod
    def fwd(u, delta, A, B, C, D, dummy, delta_bias, delta_softplus):
        batch, seq_len, d_model = u.shape
        x = torch.randn(batch, d_model, dtype=u.dtype)
        out = torch.randn_like(u)
        return out, x

    @staticmethod
    def bwd(u, delta, A, B, C, D, dummy, delta_bias, dout, x, dummy2, dummy3, delta_softplus, dummy4):
        du = torch.randn_like(u)
        ddelta = torch.randn_like(delta)
        dA = torch.randn_like(A)
        dB = torch.randn_like(B)
        dC = torch.randn_like(C)
        dD = torch.randn_like(D) if D is not None else None
        ddelta_bias = torch.randn_like(delta_bias) if delta_bias is not None else None
        return du, ddelta, dA, dB, dC, dD, ddelta_bias

class MockSelectiveScanCudaCore:
    @staticmethod
    def fwd(u, delta, A, B, C, D, delta_bias, delta_softplus, nrows):
        batch, seq_len, d_model = u.shape
        x = torch.randn(batch, d_model, dtype=u.dtype)
        out = torch.randn_like(u)
        return out, x

    @staticmethod
    def bwd(u, delta, A, B, C, D, delta_bias, dout, x, delta_softplus, nrows):
        du = torch.randn_like(u)
        ddelta = torch.randn_like(delta)
        dA = torch.randn_like(A)
        dB = torch.randn_like(B)
        dC = torch.randn_like(C)
        dD = torch.randn_like(D) if D is not None else None
        ddelta_bias = torch.randn_like(delta_bias) if delta_bias is not None else None
        return du, ddelta, dA, dB, dC, dD, ddelta_bias

# Inject mocks into global namespace
import sys
sys.modules['selective_scan_cuda'] = type(sys)('selective_scan_cuda')
sys.modules['selective_scan_cuda'].fwd = MockSelectiveScanCuda.fwd
sys.modules['selective_scan_cuda'].bwd = MockSelectiveScanCuda.bwd
sys.modules['selective_scan_cuda_core'] = type(sys)('selective_scan_cuda_core')
sys.modules['selective_scan_cuda_core'].fwd = MockSelectiveScanCudaCore.fwd
sys.modules['selective_scan_cuda_core'].bwd = MockSelectiveScanCudaCore.bwd

selective_scan_cuda = sys.modules['selective_scan_cuda']
selective_scan_cuda_core = sys.modules['selective_scan_cuda_core']

SSMODE = 'mamba_ssm'

# Construct valid arguments
batch = 2
seq_len = 8
d_model = 4
state_dim = 3
nrows_val = 1

u = torch.randn(batch, seq_len, d_model, dtype=torch.float32)
delta = torch.randn(batch, seq_len, d_model, dtype=torch.float32)
A = torch.randn(d_model, state_dim, dtype=torch.float32)
B = torch.randn(batch, seq_len, state_dim, dtype=torch.float32)
C = torch.randn(batch, seq_len, state_dim, dtype=torch.float32)
D = torch.randn(d_model, dtype=torch.float32)
delta_bias = torch.randn(d_model, dtype=torch.float32)

ARGS = {
    'u': u,
    'delta': delta,
    'A': A,
    'B': B,
    'C': C,
    'D': D,
    'delta_bias': delta_bias,
    'delta_softplus': False,
    'nrows': nrows_val,
}
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 SelectiveScan — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SelectiveScan 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
VSSM Ran
ran on a synthesized input · constructed an object; no result was computed
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("59be5ca644957b0e")
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 an instance of its own class, VSSM: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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
import math
from functools import partial

torch.manual_seed(0)

# Build a minimal VSSM-compatible module
class MinimalVSSM(nn.Module):
    def __init__(self, in_channels=4, hidden_dim=8, num_layers=2):
        super().__init__()
        self.in_channels = in_channels
        self.hidden_dim = hidden_dim
        self.num_layers = num_layers
        
        self.input_proj = nn.Linear(in_channels, hidden_dim)
        self.layers = nn.ModuleList([
            nn.Linear(hidden_dim, hidden_dim) for _ in range(num_layers)
        ])
        self.output_proj = nn.Linear(hidden_dim, in_channels)
    
    def forward(self, x):
        # x shape: (B, C, H, W)
        B, C, H, W = x.shape
        x_flat = x.flatten(2, 3).transpose(1, 2)  # (B, H*W, C)
        x = self.input_proj(x_flat)
        for layer in self.layers:
            x = layer(x)
        x = self.output_proj(x)
        x = x.transpose(1, 2).view(B, C, H, W)
        return x

# Create minimal config
config = {
    'in_channels': 4,
    'hidden_dim': 8,
    'num_layers': 2,
    'depth': 2,
    'mlp_ratio': 2,
    'drop_path': 0.1,
}

# Create input tensor
x = torch.randn(2, 4, 8, 8)

# Create model
model = MinimalVSSM(
    in_channels=config['in_channels'],
    hidden_dim=config['hidden_dim'],
    num_layers=config['num_layers']
)

ARGS = {
    'x': x,
    'model': model,
    'config': config,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type VSSM — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a VSSM 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
Permute Not yet run
unverified (harvested, no recorded run)
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("f3bea939f398ae2f")
How it ran

no invocation recorded.

VSSBlock Not yet run
unverified (harvested, no recorded run)
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("b18ae335bc5cb05c")
How it ran

no invocation recorded.

selective_scan_flatten Not yet run
unverified (harvested, no recorded run)
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("98980b013aa42f1f")
How it ran

no invocation recorded.

x_selective_scan Not yet run
unverified (harvested, no recorded run)
mkang315/PanopMamba/mmseg/models/backbones/msvssm.py
code served (permissive licence) · get_code("200638f6e4ee0cff")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 0 later) and 13 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
CrossScan
brick of this paper's assembly: model
2 other papers
2 dated earlier · 2 only as an assembly brick
2312.00752 (2023-12)
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
2506.00466 (2025-06)
M3ANet: Multi-scale and Multi-Modal Alignment Network for Brain-Assisted Target Speaker Extraction
CrossMerge
brick of this paper's assembly: model
1 other paper
1 dated earlier · 1 only as an assembly brick
2312.00752 (2023-12)
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
Permute
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2506.00466 (2025-06)
M3ANet: Multi-scale and Multi-Modal Alignment Network for Brain-Assisted Target Speaker Extraction
SelectiveScan
brick of this paper's assembly: model
1 other paper
1 dated earlier · 1 only as an assembly brick
2312.00752 (2023-12)
Mamba: Linear-Time Sequence Modeling with Selective State Spaces

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

Nuclei panoptic segmentation supports cancer diagnostics by integrating both semantic and instance segmentation of different cell types to analyze overall tissue structure and individual nuclei in histopathology images. Major challenges include detecting small objects, handling ambiguous boundaries, and addressing class imbalance. To address these issues, we propose Panop-Mamba, a novel hybrid encoder-decoder architecture that integrates Mamba and Transformer with additional featureenhanced fusion via state space modeling. We design a multiscale Mamba backbone and a State Space Model (SSM)-based fusion network to enable efficient long-range perception in pyramid features, thereby extending the pure encoder-decoder framework while facilitating information sharing across multiscale features of nuclei. The proposed SSM-based feature-enhanced fusion integrates pyramid feature networks and dynamic feature enhancement across different spatial scales, enhancing the feature representation of densely overlapping nuclei in both semantic and spatial dimensions. To the best of our knowledge, this is the first Mamba-based approach for panoptic segmentation. Additionally, we introduce alternative evaluation metrics, including image-level Panoptic Quality (iPQ), boundaryweighted PQ (wPQ), and frequency-weighted PQ (fwPQ), which are specifically designed to address the unique challenges of nuclei segmentation and thereby mitigate the potential bias inherent in vanilla PQ. Experimental evaluations on two multiclass nuclei segmentation benchmark datasets, MoNuSAC2020 and NuInsSeg, demonstrate the superiority of PanopMamba for nuclei panoptic segmentation over state-of-the-art methods. Ablation studies further validate the effectiveness of each component in improving performance. Consequently, the robustness of PanopMamba is validated across various metrics, while the distinctiveness of PQ variants is also demonstrated. Code is available at https://github.com/mkang315/PanopMamba.

For agents

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

get_harvested_code_for_paper("2601.16631")
get_code_for_paper("2601.16631")
have("2601.16631")

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

curl https://syntology.ai/api/ran/2601.16631.json

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

[![Syntology run record](https://syntology.ai/api/ran/2601.16631.svg)](https://syntology.ai/paper/2601.16631)

Connect an agent — have() is free.