SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.27552 · 2026

BLOSSOM: Block-wise Federated Learning Over Shared and Sparse Observed Modalities

Ahmed Abdelmoniem, M Pranav, Jayant Chandwani, Arnab Paul

arXiv · PDF · Open in the Atlas

Code that ran

5 of 5 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; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 5 failures of our instrument; 0 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/2603.27552.json · badge: /api/ran/2603.27552.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
DaSH-Lab-CSIS/blossom canonical (classifier label)
classifier basis: census
5 ran · 0 no recorded run
FunctionStatusWhere it lives
build_model Ran
ran; our contract draft was wrong, not the code
DaSH-Lab-CSIS/blossom/blossom/models.py
code served (permissive licence) · get_code("f97478f4b7adb5d4")
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 LateFusion. 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
from torch import nn

torch.manual_seed(0)

# Define small encoder modules for each modality
class TextEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(10, 8)
    
    def forward(self, x):
        return self.linear(x)

class ImageEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(16, 8)
    
    def forward(self, x):
        return self.linear(x)

class AudioEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(12, 8)
    
    def forward(self, x):
        return self.linear(x)

# Define a fusion module that takes a dict of embeddings and returns fused representation
class FusionModule(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(24, 16)  # 3 modalities * 8 dim each = 24
    
    def forward(self, embeddings):
        # embeddings is a dict with keys: 'text', 'image', 'audio'
        # Concatenate all embeddings
        concat = torch.cat(list(embeddings.values()), dim=1)
        return self.linear(concat)

# Define a head module for final prediction
class HeadModule(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(16, 4)
    
    def forward(self, x):
        return self.linear(x)

# Create encoder dictionary
encoders_dict = {
    'text': TextEncoder(),
    'image': ImageEncoder(),
    'audio': AudioEncoder()
}

# Create fusion and head modules
fusion_module = FusionModule()
head_module = HeadModule()

ARGS = {
    'encoders_dict': encoders_dict,
    'fusion_module': fusion_module,
    'head_module': head_module
}
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 LateFusion — a LateFusion output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a LateFusion 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
extract_modality_metrics_from_client_data Ran
ran; our contract draft was wrong, not the code
DaSH-Lab-CSIS/blossom/blossom/utils.py
code served (permissive licence) · get_code("a4891589387b85fd")
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 pandas as pd
import numpy as np

# Create sample dataframes with client names containing modality information
df1 = pd.DataFrame({
    'client': ['client_1(text)', 'client_2(text)', 'client_3(image)', 'client_4(image)'],
    'round': [1, 1, 1, 1],
    'accuracy': [0.85, 0.87, 0.82, 0.84],
    'loss': [0.15, 0.13, 0.18, 0.16]
})

df2 = pd.DataFrame({
    'client': ['client_1(text)', 'client_2(text)', 'client_3(image)', 'client_4(image)'],
    'round': [2, 2, 2, 2],
    'accuracy': [0.88, 0.89, 0.85, 0.87],
    'loss': [0.12, 0.11, 0.15, 0.13]
})

df3 = pd.DataFrame({
    'client': ['client_5(audio)', 'client_6(audio)'],
    'round': [1, 1],
    'accuracy': [0.80, 0.81],
    'loss': [0.20, 0.19]
})

client_dataframes = [df1, df2, df3]

ARGS = {
    'client_dataframes': client_dataframes
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — 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
mask_modalities Ran
ran; our contract draft was wrong, not the code
DaSH-Lab-CSIS/blossom/blossom/task.py
code served (permissive licence) · get_code("0149d6d3b64cc15d")
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 torch

torch.manual_seed(0)

# Create a minimal batch with tensor values and non-modality keys
batch = {
    "audio": torch.randn(2, 4),
    "vision": torch.randn(2, 4),
    "text": torch.randn(2, 4),
    "metadata": torch.randn(2, 3),
}

# Define which modalities are present
modalities = ("audio", "vision")

# Define all possible modalities in the system
all_modalities = ["audio", "vision", "text"]

# Set device to CPU
device = torch.device("cpu")

ARGS = {
    "batch": batch,
    "modalities": modalities,
    "all_modalities": all_modalities,
    "device": device,
}
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 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
parse_dict_with_tuple_keys Ran
ran; our contract draft was wrong, not the code
DaSH-Lab-CSIS/blossom/blossom/utils.py
code served (permissive licence) · get_code("a3d02a5475c08b8b")
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 torch
import numpy as np
from typing import Dict, Tuple

torch.manual_seed(0)

# Construct a valid source dictionary with string keys representing tuples
src_dict = {
    "audio_video": 10,
    "text_image": 20,
    "audio_text_video": 30,
    "modality_one": 5
}

ARGS = {
    "src_dict": src_dict
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — 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
prepare_lines_data_from_modality_metrics Ran
ran; our contract draft was wrong, not the code
DaSH-Lab-CSIS/blossom/blossom/utils.py
code served (permissive licence) · get_code("06cfc69fb091fa68")
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 numpy as np

# Construct modality_metrics dictionary
modality_metrics = {
    'audio': {
        'accuracy': {
            0: [0.75, 0.80, 0.78],
            1: [0.82, 0.85, 0.83],
            2: [0.88, 0.90, 0.89]
        },
        'loss': {
            0: [0.5, 0.45],
            1: [0.3, 0.28]
        }
    },
    'vision': {
        'accuracy': {
            0: [0.70, 0.72],
            1: [0.80, 0.81, 0.82],
            2: [0.85, 0.87]
        },
        'loss': {
            0: [0.6, 0.55, 0.52],
            1: [0.4, 0.35]
        }
    },
    'text': {
        'accuracy': {
            0: [0.65, 0.68, 0.67],
            1: [0.75, 0.78],
            2: [0.82, 0.85, 0.83]
        }
    }
}

metric = 'accuracy'

ARGS = {
    'modality_metrics': modality_metrics,
    'metric': metric
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 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 5 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

Multimodal federated learning (FL) is essential for real-world applications such as autonomous systems and healthcare, where data is distributed across heterogeneous clients with varying and often missing modalities. However, most existing FL approaches assume uniform modality availability, limiting their applicability in practice. We introduce BLOSSOM, a taskagnostic framework for multimodal FL designed to operate under shared and sparsely observed modality conditions. BLOSSOM supports clients with arbitrary modality subsets and enables flexible sharing of model components. To address client and task heterogeneity, we propose a block-wise aggregation strategy that selectively aggregates shared components while keeping task-specific blocks private, enabling partial personalization. We evaluate BLOSSOM on multiple diverse multimodal datasets and analyse the effects of missing modalities and personalization. Our results show that block-wise personalization significantly improves performance, particularly in settings with severe modality sparsity. In modality-incomplete scenarios, BLOSSOM achieves an average performance gain of 18.7% over full-model aggregation, while in modality-exclusive settings the gain increases to 37.7%, highlighting the importance of block-wise learning for practical multimodal FL systems.

For agents

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

get_harvested_code_for_paper("2603.27552")
get_code_for_paper("2603.27552")
have("2603.27552")

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

curl https://syntology.ai/api/ran/2603.27552.json

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

[![Syntology run record](https://syntology.ai/api/ran/2603.27552.svg)](https://syntology.ai/paper/2603.27552)

Connect an agent — have() is free.